title
Photogrammetry I - 01 - Introduction (2015)

description
Photogrammetry I Course, Chapter: Introduction Slides: Send me an email and you can download the slides. This lecture is part of the Photogrammetry I course at BSc level taught by Cyrill Stachniss at the University of Bonn, Germany in the summer term 2015. Slide Information * The slides have been created by Cyrill Stachniss as part of the photogrammetry and/or robotics courses. * I tried to acknowledge all people who contributed images or videos. In case I made a mistake or missed someone, please let me know. * The photogrammetry material heavily relies on the very well written lecture notes by Wolfgang Förstner and the Photogrammetric Computer Vision book by Förstner & Wrobel. * Parts of the robotics material, especially the algorithms, stems from the great Probabilistic Robotics book by Thrun, Burgard and Fox. * Feel free to use the course material, please send me an email to receive the Powerpoint files.

detail
{'title': 'Photogrammetry I - 01 - Introduction (2015)', 'heatmap': [{'end': 1226.451, 'start': 1193.993, 'weight': 0.748}, {'end': 1312.447, 'start': 1247.297, 'weight': 1}, {'end': 1397.846, 'start': 1361.746, 'weight': 0.715}, {'end': 1679.766, 'start': 1646.545, 'weight': 0.803}], 'summary': 'Introduces photogrammetry, discussing its principles, applications, and estimation of geometric properties based on images. it also explores the use of camera sensors for 3d geometry estimation, the benefits of contact-free sensing, photogrammetry fundamentals, innovative visual perception, and applications and techniques in 3d reconstruction, including city mapping and cultural heritage preservation.', 'chapters': [{'end': 219.337, 'segs': [{'end': 42.767, 'src': 'embed', 'start': 0.929, 'weight': 1, 'content': [{'end': 8.553, 'text': 'Okay, so the goal for the course today, after having clarified all the introductory material,', 'start': 0.929, 'duration': 7.624}, {'end': 11.574, 'text': 'is actually to stimulate your interest in photogrammetry.', 'start': 8.553, 'duration': 3.021}, {'end': 21.519, 'text': 'So we will not go into a lot of technical details today, but I just would like to give you a very very brief introduction to some basic techniques.', 'start': 12.074, 'duration': 9.445}, {'end': 26.061, 'text': 'or some basic principles, not even techniques, and then show applications.', 'start': 22.139, 'duration': 3.922}, {'end': 30.302, 'text': "So you'll see a couple of nice images, nice videos today, just to stimulate your interest.", 'start': 26.141, 'duration': 4.161}, {'end': 32.183, 'text': "And that's kind of my goal for today.", 'start': 30.742, 'duration': 1.441}, {'end': 37.545, 'text': 'And the first question to answer is actually what is photogrammetry, or where does the word come from?', 'start': 33.043, 'duration': 4.502}, {'end': 42.767, 'text': "It's a Greek word and it has something to do with light, photos of light, and it has something to do with measuring.", 'start': 37.605, 'duration': 5.162}], 'summary': 'The goal is to stimulate interest in photogrammetry with a brief introduction to basic principles and applications.', 'duration': 41.838, 'max_score': 0.929, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY929.jpg'}, {'end': 113.544, 'src': 'embed', 'start': 68.8, 'weight': 0, 'content': [{'end': 71.281, 'text': 'But it can also mean a semantic interpretation.', 'start': 68.8, 'duration': 2.481}, {'end': 74.541, 'text': 'So we want to identify what we actually see there.', 'start': 71.621, 'duration': 2.92}, {'end': 81.943, 'text': "So that some box-like object that we see is actually not just a stupid box, but it's, for example, a car or a building.", 'start': 74.722, 'duration': 7.221}, {'end': 83.884, 'text': 'This is kind of the idea of semantic.', 'start': 82.564, 'duration': 1.32}, {'end': 87.745, 'text': 'So we want to understand what we actually see or analyze the scene.', 'start': 84.464, 'duration': 3.281}, {'end': 93.85, 'text': 'So, if you look to the definitions, one of the definitions we find about photogrammetry.', 'start': 88.826, 'duration': 5.024}, {'end': 100.194, 'text': "it's the estimation of the geometric and semantic properties of objects based on images or observations from similar sensors.", 'start': 93.85, 'duration': 6.344}, {'end': 103.797, 'text': 'We said, okay, geometry should be pretty clear.', 'start': 102.016, 'duration': 1.781}, {'end': 110.182, 'text': 'Semantic, maybe not perfectly clear yet, but simply see that as names.', 'start': 105.198, 'duration': 4.984}, {'end': 113.544, 'text': 'So we can give names to objects, like car, building.', 'start': 110.502, 'duration': 3.042}], 'summary': 'Identifying objects in images with semantic interpretation for photogrammetry.', 'duration': 44.744, 'max_score': 68.8, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY68800.jpg'}, {'end': 234.984, 'src': 'embed', 'start': 201.168, 'weight': 3, 'content': [{'end': 205.03, 'text': 'And what the laser scanner basically does, it emits a laser signal.', 'start': 201.168, 'duration': 3.862}, {'end': 206.771, 'text': 'The signal is reflected from that wall.', 'start': 205.27, 'duration': 1.501}, {'end': 214.815, 'text': 'And it simply measures the time it takes for the light to travel to the object and to travel back to a sensor in here.', 'start': 207.531, 'duration': 7.284}, {'end': 216.996, 'text': 'And then it can actually calculate the distance.', 'start': 215.475, 'duration': 1.521}, {'end': 219.337, 'text': 'What else do we have?', 'start': 218.717, 'duration': 0.62}, {'end': 234.984, 'text': 'Infrared is also definitely something very similar to cameras, even sensors of cameras, and in infrared that can receive infrared light,', 'start': 225.979, 'duration': 9.005}], 'summary': 'Laser scanner measures distance using reflected light. infrared sensors detect infrared light.', 'duration': 33.816, 'max_score': 201.168, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY201168.jpg'}], 'start': 0.929, 'title': 'Photogrammetry', 'summary': 'Introduces and discusses photogrammetry, covering basic principles, applications, and estimation of geometric and semantic properties of objects based on images or similar sensors.', 'chapters': [{'end': 67.655, 'start': 0.929, 'title': 'Introduction to photogrammetry', 'summary': 'Introduces the concept of photogrammetry, aiming to stimulate interest by providing a brief overview of basic principles and applications, such as the measurement of objects and interpretation of geometric data using photographs.', 'duration': 66.726, 'highlights': ['Photogrammetry is introduced to stimulate interest by providing a brief overview of basic principles and applications, such as the measurement of objects and interpretation of geometric data using photographs.', 'The course aims to stimulate interest in photogrammetry through the presentation of compelling images and videos to the audience.', 'Photogrammetry involves measuring and interpreting the data obtained from photographs to estimate the geometry of objects.', "The word 'photogrammetry' originates from the Greek words for 'light' and 'measuring', reflecting the process of measuring with light, particularly through photographs."]}, {'end': 219.337, 'start': 68.8, 'title': 'Understanding photogrammetry and semantic interpretation', 'summary': 'Discusses the concept of photogrammetry, which involves estimating geometric and semantic properties of objects based on images or observations from similar sensors, such as radar, satellites, and laser scanning.', 'duration': 150.537, 'highlights': ['Photogrammetry involves estimating geometric and semantic properties of objects based on images or observations from similar sensors, such as radar, satellites, and laser scanning.', 'Semantic interpretation refers to giving names to objects, like car, building, and understanding what is seen in the scene.', 'Laser scanning in photogrammetry emits a laser signal, measures the time it takes for the light to travel to the object and back to the sensor, and calculates the distance.']}], 'duration': 218.408, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY929.jpg', 'highlights': ['Photogrammetry involves estimating geometric and semantic properties of objects based on images or similar sensors.', 'The course aims to stimulate interest in photogrammetry through the presentation of compelling images and videos to the audience.', "The word 'photogrammetry' originates from the Greek words for 'light' and 'measuring', reflecting the process of measuring with light, particularly through photographs.", 'Laser scanning in photogrammetry emits a laser signal, measures the time it takes for the light to travel to the object and back to the sensor, and calculates the distance.', 'Semantic interpretation refers to giving names to objects, like car, building, and understanding what is seen in the scene.']}, {'end': 850.903, 'segs': [{'end': 314.619, 'src': 'embed', 'start': 277.573, 'weight': 0, 'content': [{'end': 283.178, 'text': 'Or we can actually move the camera and get multiple images of the same scene from slightly different viewpoints.', 'start': 277.573, 'duration': 5.605}, {'end': 290.085, 'text': 'And this allows us, for example, to estimate 3D geometry or estimate changes over time in the scene.', 'start': 283.839, 'duration': 6.246}, {'end': 295.363, 'text': 'These are all things that we can actually use for video cameras or smartphones.', 'start': 290.759, 'duration': 4.604}, {'end': 296.984, 'text': 'Smartphones may have other sensors as well.', 'start': 295.383, 'duration': 1.601}, {'end': 301.949, 'text': 'Typically, you have an inertial measurement unit in there which tells you something about your attitude.', 'start': 297.545, 'duration': 4.404}, {'end': 304.03, 'text': 'gives you information about your attitude quite well.', 'start': 301.949, 'duration': 2.081}, {'end': 306.733, 'text': 'You can actually exploit that in the estimation processes.', 'start': 304.571, 'duration': 2.162}, {'end': 314.619, 'text': 'So all this can be seen as part of photogrammetry, although traditional cameras are kind of in the main focus, at least of this course here today.', 'start': 307.273, 'duration': 7.346}], 'summary': 'Utilizing multiple images for 3d estimation and scene changes; applicable to video cameras and smartphones with additional sensors.', 'duration': 37.046, 'max_score': 277.573, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY277573.jpg'}, {'end': 375.64, 'src': 'embed', 'start': 355.177, 'weight': 2, 'content': [{'end': 365.36, 'text': 'I can also infer something about the geometry of the scene by once taking a picture of a person sitting there on the back from this location and from this location.', 'start': 355.177, 'duration': 10.183}, {'end': 368.741, 'text': 'I can actually, by triangulation, estimate how far that person is away.', 'start': 365.84, 'duration': 2.901}, {'end': 375.64, 'text': 'So it allows me to get some information about the location of the object or a 3D reconstruction of the scene or the objects.', 'start': 369.755, 'duration': 5.885}], 'summary': 'Estimating distance using triangulation for 3d reconstruction.', 'duration': 20.463, 'max_score': 355.177, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY355177.jpg'}, {'end': 445.976, 'src': 'embed', 'start': 420.598, 'weight': 1, 'content': [{'end': 426.244, 'text': 'This is important for a lot of things where we want to monitor tasks and want to get interpretation of the scenes.', 'start': 420.598, 'duration': 5.646}, {'end': 431.049, 'text': 'Or very popular at the moment is things in the context of autonomous driving.', 'start': 426.644, 'duration': 4.405}, {'end': 433.952, 'text': 'So we have cars that can actually drive autonomously.', 'start': 431.749, 'duration': 2.203}, {'end': 438.113, 'text': "But these cars actually need to understand what's going on in front of the scene.", 'start': 435.032, 'duration': 3.081}, {'end': 445.976, 'text': "So the classical example is if you have kids playing on the sidewalk, it's a risk that this person actually enters the street.", 'start': 438.814, 'duration': 7.162}], 'summary': 'Monitoring tasks and interpreting scenes is crucial, particularly in autonomous driving, where cars need to understand and react to their surroundings to ensure safety.', 'duration': 25.378, 'max_score': 420.598, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY420598.jpg'}, {'end': 568.269, 'src': 'embed', 'start': 538.962, 'weight': 5, 'content': [{'end': 542.745, 'text': 'So this process of having images and inferring something about the scene.', 'start': 538.962, 'duration': 3.783}, {'end': 547.85, 'text': 'This is something which is, again, photogrammetry is one of the first disciplines addressing this problem.', 'start': 543.486, 'duration': 4.364}, {'end': 551.935, 'text': 'over the last twenty years, the community in computer vision,', 'start': 548.712, 'duration': 3.223}, {'end': 558.962, 'text': 'which is a branch which typically started from computer science or from electrical engineering, is actually solving very, very similar tasks,', 'start': 551.935, 'duration': 7.027}, {'end': 568.269, 'text': "maybe in a slightly different objective function not always measuring per se, but scene interpretation, change, detection, analyzing what's going on.", 'start': 558.962, 'duration': 9.307}], 'summary': 'Photogrammetry and computer vision address scene interpretation and change detection over the past 20 years.', 'duration': 29.307, 'max_score': 538.962, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY538962.jpg'}, {'end': 816.023, 'src': 'embed', 'start': 776.749, 'weight': 6, 'content': [{'end': 783.374, 'text': 'And last but not least, especially the aspect of state estimation robotics is very, very closely related to photogrammetry.', 'start': 776.749, 'duration': 6.625}, {'end': 791.218, 'text': "So the question now, what are the advantages of using photogrammetric techniques? The first thing I said before, it's contact-free sensing.", 'start': 784.414, 'duration': 6.804}, {'end': 793.879, 'text': "So we don't need to touch the object.", 'start': 792.318, 'duration': 1.561}, {'end': 816.023, 'text': 'Why may it be an advantage to do contact-free sensing? Any ideas? Why is it a good idea to do contact-free sensing? Yeah? Perfectly fine.', 'start': 794.92, 'duration': 21.103}], 'summary': 'State estimation in robotics is related to photogrammetry, offering contact-free sensing and eliminating the need to touch the object.', 'duration': 39.274, 'max_score': 776.749, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY776749.jpg'}], 'start': 225.979, 'title': 'Camera sensors and photogrammetry', 'summary': 'Discusses the use of cameras and sensors in infrared and mobile phones for estimating 3d geometry and scene changes, emphasizing the importance of scene interpretation for tasks like autonomous driving and the overlap of photogrammetry with computer graphics, computer vision, robotics, and remote sensing, underlining the contact-free sensing advantage and the influence of computer vision on photogrammetry.', 'chapters': [{'end': 455.96, 'start': 225.979, 'title': 'Camera sensors and photogrammetry', 'summary': 'Discusses the use of cameras and sensors in infrared and mobile phones for capturing multiple images and exploiting them to estimate 3d geometry and changes in scenes, emphasizing the importance of scene interpretation for tasks like autonomous driving.', 'duration': 229.981, 'highlights': ['The importance of capturing multiple images with camera sensors and exploiting them to estimate 3D geometry and changes in scenes, such as in smartphones recording videos at 25-30 frames per second.', 'The significance of scene interpretation for tasks like autonomous driving, where understanding the scene is crucial for safe driving, such as detecting risks like kids playing on the sidewalk and ensuring their safety.', 'The ability to estimate the position where an image was taken and infer the geometry of the scene by triangulation, enabling 3D reconstruction and object recognition.', 'The utilization of inertial measurement unit in smartphones to provide information about attitude, which can be exploited in estimation processes.', 'The concept of generating a description of a scene, such as recognizing people entering a building, which is crucial for tasks like monitoring and autonomous driving.']}, {'end': 850.903, 'start': 455.96, 'title': 'Photogrammetry and its overlapping disciplines', 'summary': 'Highlights the overlap of photogrammetry with computer graphics, computer vision, robotics, and remote sensing, emphasizing the contact-free sensing advantage and the influence of computer vision on photogrammetry.', 'duration': 394.943, 'highlights': ["Photogrammetry's overlap with computer graphics, computer vision, robotics, and remote sensing The chapter emphasizes the substantial overlap of photogrammetry with computer graphics, computer vision, robotics, and remote sensing, showcasing the convergence of these disciplines.", 'Advantages of contact-free sensing in photogrammetry The advantage of contact-free sensing in photogrammetry is highlighted, emphasizing its relevance in situations where objects are inaccessible, hazardous, or fragile, such as in the case of mountains, volcanoes, toxic materials, or extreme temperature environments.', 'Influence of computer vision on photogrammetry The chapter indicates that the development of photogrammetry has been influenced by the field of computer vision, showcasing the impact of computer vision on the advancements in photogrammetric techniques.']}], 'duration': 624.924, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY225979.jpg', 'highlights': ['The importance of capturing multiple images with camera sensors and exploiting them to estimate 3D geometry and changes in scenes, such as in smartphones recording videos at 25-30 frames per second.', 'The significance of scene interpretation for tasks like autonomous driving, where understanding the scene is crucial for safe driving, such as detecting risks like kids playing on the sidewalk and ensuring their safety.', 'The ability to estimate the position where an image was taken and infer the geometry of the scene by triangulation, enabling 3D reconstruction and object recognition.', 'The utilization of inertial measurement unit in smartphones to provide information about attitude, which can be exploited in estimation processes.', 'The concept of generating a description of a scene, such as recognizing people entering a building, which is crucial for tasks like monitoring and autonomous driving.', "Photogrammetry's overlap with computer graphics, computer vision, robotics, and remote sensing The chapter emphasizes the substantial overlap of photogrammetry with computer graphics, computer vision, robotics, and remote sensing, showcasing the convergence of these disciplines.", 'Advantages of contact-free sensing in photogrammetry The advantage of contact-free sensing in photogrammetry is highlighted, emphasizing its relevance in situations where objects are inaccessible, hazardous, or fragile, such as in the case of mountains, volcanoes, toxic materials, or extreme temperature environments.', 'Influence of computer vision on photogrammetry The chapter indicates that the development of photogrammetry has been influenced by the field of computer vision, showcasing the impact of computer vision on the advancements in photogrammetric techniques.']}, {'end': 1420.161, 'segs': [{'end': 876.356, 'src': 'embed', 'start': 853.612, 'weight': 0, 'content': [{'end': 863.142, 'text': 'Yep, so definitely if something is very, very small, maybe easier to handle that with a camera or very sensitive material.', 'start': 853.612, 'duration': 9.53}, {'end': 869.388, 'text': "These are kind of the main advantages of this contact-free sensing, which even pointed out more, that's great.", 'start': 864.203, 'duration': 5.185}, {'end': 876.356, 'text': 'The other thing why this is nice is actually that we can cover large areas quite easily.', 'start': 871.351, 'duration': 5.005}], 'summary': 'Contact-free sensing offers advantages for handling small items and covering large areas easily.', 'duration': 22.744, 'max_score': 853.612, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY853612.jpg'}, {'end': 980.779, 'src': 'embed', 'start': 919.154, 'weight': 1, 'content': [{'end': 924.357, 'text': 'very close and get a quite detailed view of what I see, but then cover only a small area.', 'start': 919.154, 'duration': 5.203}, {'end': 928.86, 'text': 'And I have all this flexibility within the process without at least the algorithms not caring.', 'start': 924.397, 'duration': 4.463}, {'end': 937.3, 'text': "So we may choose a slightly different hardware, but the algorithms for processing this data It doesn't make any difference.", 'start': 929.54, 'duration': 7.76}, {'end': 940.122, 'text': 'We can actually do 2D sensing and 3D sensing.', 'start': 938, 'duration': 2.122}, {'end': 948.048, 'text': 'So one of the things is if I take an image of a scene and actually project a 3D scene to a 2D world,', 'start': 940.302, 'duration': 7.746}, {'end': 950.59, 'text': "that's a problem from which you will suffer in the next year.", 'start': 948.048, 'duration': 2.542}, {'end': 958.417, 'text': 'So actually you lose information, you lose the depth information if you project the 3D world on a 2D image plane.', 'start': 951.571, 'duration': 6.846}, {'end': 966.554, 'text': 'But if you take multiples of those images, we can actually try to reconstruct the 3D geometry of the scene.', 'start': 960.404, 'duration': 6.15}, {'end': 969.658, 'text': "We can't do it with a single image, but we can do that with multiple images.", 'start': 966.674, 'duration': 2.984}, {'end': 974.987, 'text': 'It allows us to do 2D sensing activities and even 3D sensing activities if we have, at least if we have multiple images.', 'start': 969.678, 'duration': 5.309}, {'end': 978.017, 'text': 'There are a couple of other really important things.', 'start': 976.135, 'duration': 1.882}, {'end': 980.779, 'text': 'Dynamics can be actually recorded quite easily.', 'start': 978.417, 'duration': 2.362}], 'summary': 'Flexibility in process with algorithms for 2d and 3d sensing, reconstructing 3d with multiple images.', 'duration': 61.625, 'max_score': 919.154, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY919154.jpg'}, {'end': 1096.129, 'src': 'embed', 'start': 1059.934, 'weight': 3, 'content': [{'end': 1068.839, 'text': "Also something which is a big advantage, especially if it's very hard to obtain the data, or I have only one chance to obtain the data,", 'start': 1059.934, 'duration': 8.905}, {'end': 1071.58, 'text': 'and maybe something I made a mistake in the measuring process.', 'start': 1068.839, 'duration': 2.741}, {'end': 1074.322, 'text': 'I can often recover that by going back to the original images.', 'start': 1071.58, 'duration': 2.742}, {'end': 1079.097, 'text': 'Especially since cameras are used for a lot of different applications.', 'start': 1076.035, 'duration': 3.062}, {'end': 1084.621, 'text': 'cameras are comparably cheap, because every one of us has a cell phone with a camera nearly every one of us.', 'start': 1079.097, 'duration': 5.524}, {'end': 1086.282, 'text': 'So cameras are a mass product.', 'start': 1085.001, 'duration': 1.281}, {'end': 1090.725, 'text': 'And this makes them actually comparably cheap compared to, for example, other sensors.', 'start': 1086.802, 'duration': 3.923}, {'end': 1095.448, 'text': 'You can do real-time processing nowadays or even fully automatic data processing.', 'start': 1090.745, 'duration': 4.703}, {'end': 1096.129, 'text': "So it's all great.", 'start': 1095.468, 'duration': 0.661}], 'summary': 'Cameras are advantageous due to mass production, affordability, and ability for real-time processing.', 'duration': 36.195, 'max_score': 1059.934, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1059934.jpg'}, {'end': 1146.786, 'src': 'embed', 'start': 1121.394, 'weight': 5, 'content': [{'end': 1128.137, 'text': "I actually can measure only at 50% of the time if I can't employ an artificial light source.", 'start': 1121.394, 'duration': 6.743}, {'end': 1130.678, 'text': 'So light is needed, which can be a real disadvantage.', 'start': 1128.637, 'duration': 2.041}, {'end': 1133.559, 'text': 'Depending on the application that you have, this can be a limiting factor.', 'start': 1130.718, 'duration': 2.841}, {'end': 1136.698, 'text': 'You also need to take care about occlusions.', 'start': 1135.277, 'duration': 1.421}, {'end': 1144.024, 'text': "So one object occluding another object that you can't see it, which can actually make all the measurement process quite hard.", 'start': 1137.319, 'duration': 6.705}, {'end': 1146.786, 'text': 'The other tensing modalities like radar was one example.', 'start': 1144.064, 'duration': 2.722}], 'summary': 'Measurement accuracy is limited to 50% without artificial light, posing a disadvantage and potential limits on application. occlusions and other sensing modalities, such as radar, also present challenges.', 'duration': 25.392, 'max_score': 1121.394, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1121394.jpg'}, {'end': 1231.432, 'src': 'heatmap', 'start': 1185.666, 'weight': 6, 'content': [{'end': 1190.21, 'text': "For example, you can't estimate the distance, at least from a single image.", 'start': 1185.666, 'duration': 4.544}, {'end': 1192.472, 'text': 'If you have multiple images, we can actually do a better job in here.', 'start': 1190.23, 'duration': 2.242}, {'end': 1200.559, 'text': 'And there also may be other techniques which are actually more accurate than photogrammetric techniques.', 'start': 1193.993, 'duration': 6.566}, {'end': 1208.027, 'text': 'Also in terms of the relative precision, we can actually, the nice, so this is actually an advantage or a disadvantage.', 'start': 1202.386, 'duration': 5.641}, {'end': 1213.928, 'text': "So I have a relative precision, and I'm always kind of in that area over here.", 'start': 1208.067, 'duration': 5.861}, {'end': 1218.169, 'text': 'If I do microscopy, I can actually investigate very, very small objects.', 'start': 1214.408, 'duration': 3.761}, {'end': 1226.451, 'text': "If I go to traditional photogrammetric techniques, ranging, let's say, from millimeters or centimeters, let's say, tens or hundreds of meters,", 'start': 1220.35, 'duration': 6.101}, {'end': 1231.432, 'text': 'or if I go to remote sensing applications, where I cover areas of kilometers or square kilometers of size.', 'start': 1226.451, 'duration': 4.981}], 'summary': 'Multiple images improve distance estimation. microscopy for small objects, photogrammetry for larger areas.', 'duration': 45.766, 'max_score': 1185.666, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1185666.jpg'}, {'end': 1312.447, 'src': 'heatmap', 'start': 1247.297, 'weight': 1, 'content': [{'end': 1252.399, 'text': 'Although there are techniques which may be more precise than those photogrammetric techniques.', 'start': 1247.297, 'duration': 5.102}, {'end': 1257.581, 'text': 'And one of the things already mentioned that cameras actually measure directions.', 'start': 1253.979, 'duration': 3.602}, {'end': 1264.047, 'text': 'So if this is a real 3D world, this would be the projection center of my camera.', 'start': 1258.942, 'duration': 5.105}, {'end': 1266.07, 'text': 'It actually generates an image which is upside down.', 'start': 1264.067, 'duration': 2.003}, {'end': 1275.359, 'text': 'What the camera basically does, every of these pixels here in this image corresponds to a ray that goes through the 3D world.', 'start': 1266.65, 'duration': 8.709}, {'end': 1281.526, 'text': 'So every point which lies on that green ray will actually be mapped to this pixel.', 'start': 1276.36, 'duration': 5.166}, {'end': 1289.323, 'text': "So if there's an object here and here, let's say a bird flying through the scene, it will map the bird, simply the closest one.", 'start': 1282.779, 'duration': 6.544}, {'end': 1293.886, 'text': 'And I have no idea if this is the bird here, an object over here, or the top of the mountain.', 'start': 1289.363, 'duration': 4.523}, {'end': 1299.189, 'text': 'So cameras are just devices which measure directions, nothing else.', 'start': 1294.986, 'duration': 4.203}, {'end': 1306.473, 'text': 'However, if you take now two images and I know from which locations these images have been taken,', 'start': 1301.13, 'duration': 5.343}, {'end': 1312.447, 'text': "let's say one camera sits here and one camera sits here, and I know the correspondences.", 'start': 1306.473, 'duration': 5.974}], 'summary': 'Cameras measure directions, mapping objects to pixels in 3d world.', 'duration': 65.15, 'max_score': 1247.297, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1247297.jpg'}, {'end': 1397.846, 'src': 'heatmap', 'start': 1361.746, 'weight': 0.715, 'content': [{'end': 1365.869, 'text': 'OK, in terms of the basic principle, what lies behind that? So we have a contract-free measurement.', 'start': 1361.746, 'duration': 4.123}, {'end': 1369.051, 'text': 'And what we basically do, we measure intensities of photons.', 'start': 1366.149, 'duration': 2.902}, {'end': 1372.158, 'text': 'Photon is kind of the smallest quantity of light.', 'start': 1370.077, 'duration': 2.081}, {'end': 1379.58, 'text': 'And what we basically measure is the intensities, which are basically counting photons of a given wavelength.', 'start': 1373.318, 'duration': 6.262}, {'end': 1384.082, 'text': "And this, we'll go into the details during this course.", 'start': 1380.8, 'duration': 3.282}, {'end': 1395.225, 'text': 'And what basically every pixel in your camera does, it basically counts the number of photons that this small sensor detects in a given time frame.', 'start': 1384.582, 'duration': 10.643}, {'end': 1397.846, 'text': 'And this gives you the intensity.', 'start': 1396.726, 'duration': 1.12}], 'summary': 'Measuring photon intensities in contract-free measurement for camera pixels.', 'duration': 36.1, 'max_score': 1361.746, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1361746.jpg'}], 'start': 853.612, 'title': 'Contact-free sensing and photogrammetric techniques', 'summary': 'Explores the benefits of contact-free sensing, such as large area coverage, flexibility in resolution, 2d and 3d sensing capabilities, and real-time processing, and discusses challenges in photogrammetric techniques, including limitations in 3d to 2d projection and estimating distances and correspondences in image processing.', 'chapters': [{'end': 1096.129, 'start': 853.612, 'title': 'Advantages of contact-free sensing', 'summary': 'Discusses the advantages of contact-free sensing, including large area coverage, flexibility in resolution, 2d and 3d sensing capabilities, ease of interpretation by humans, and the ability to recover data from original images, with the added benefits of cost-effectiveness and real-time processing.', 'duration': 242.517, 'highlights': ['Contact-free sensing allows for large area coverage, making data acquisition very easy. The ability to cover large areas easily with cameras, even acquiring dense coverage of the scene through video cameras, provides a significant advantage in data acquisition.', 'Flexibility in resolution enables coarse or detailed views without impacting the processing algorithms. The flexibility to choose resolution based on distance from the object, without affecting the processing algorithms, is a key advantage.', '2D and 3D sensing capabilities are facilitated by multiple images, allowing for reconstructing the 3D geometry of a scene. The capability to perform both 2D and 3D sensing activities through the use of multiple images is highlighted, enabling the reconstruction of 3D scene geometry.', 'Ease of interpretation by humans and the ability to recover data from original images are emphasized as major advantages. The ease of human interpretation and the ability to recover data by going back to original images are noted as key advantages, especially in cases of measurement errors.', 'Cost-effectiveness and real-time processing are highlighted as additional benefits of using cameras for data acquisition. The cost-effectiveness and real-time processing capabilities of using cameras for data acquisition are highlighted, particularly due to their widespread availability and low cost.']}, {'end': 1420.161, 'start': 1096.149, 'title': 'Challenges of photogrammetric techniques', 'summary': 'Discusses the advantages and disadvantages of photogrammetric techniques, highlighting the need for light, limitations in 3d to 2d projection, and the challenge of estimating distances and correspondences in image processing.', 'duration': 324.012, 'highlights': ['The need for light is a significant disadvantage, limiting the monitoring capability to around 50% of the time if artificial light sources cannot be employed. Light is a crucial requirement for photogrammetric techniques, and the inability to use artificial light sources can restrict monitoring to only 50% of the time.', 'The challenge of estimating distances and correspondences in image processing, especially in single-image scenarios, poses a significant limitation. Estimating distances and correspondences in single-image scenarios is challenging and limits the precision of photogrammetric techniques.', 'The limitation in 3D to 2D projection results in the loss of information and makes estimating distances difficult, especially in the absence of size references. The transition from the 3D world to the 2D world leads to the loss of information and complicates distance estimation, particularly without size references.']}], 'duration': 566.549, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY853612.jpg', 'highlights': ['Contact-free sensing allows for large area coverage, making data acquisition very easy.', 'Flexibility in resolution enables coarse or detailed views without impacting the processing algorithms.', '2D and 3D sensing capabilities are facilitated by multiple images, allowing for reconstructing the 3D geometry of a scene.', 'Ease of interpretation by humans and the ability to recover data from original images are emphasized as major advantages.', 'Cost-effectiveness and real-time processing are highlighted as additional benefits of using cameras for data acquisition.', 'The need for light is a significant disadvantage, limiting the monitoring capability to around 50% of the time if artificial light sources cannot be employed.', 'The challenge of estimating distances and correspondences in image processing, especially in single-image scenarios, poses a significant limitation.', 'The limitation in 3D to 2D projection results in the loss of information and makes estimating distances difficult, especially in the absence of size references.']}, {'end': 1764.643, 'segs': [{'end': 1476.643, 'src': 'embed', 'start': 1450.665, 'weight': 0, 'content': [{'end': 1455.866, 'text': 'And every ray from the world passes through this pinhole.', 'start': 1450.665, 'duration': 5.201}, {'end': 1458.692, 'text': 'at those points over here.', 'start': 1457.611, 'duration': 1.081}, {'end': 1461.934, 'text': 'So what I see actually is get the tree upside down in here.', 'start': 1458.732, 'duration': 3.202}, {'end': 1465.496, 'text': "And the things that, oh sorry, they don't intersect here.", 'start': 1463.334, 'duration': 2.162}, {'end': 1466.176, 'text': 'That was wrong.', 'start': 1465.656, 'duration': 0.52}, {'end': 1467.417, 'text': 'Should be like this.', 'start': 1466.797, 'duration': 0.62}, {'end': 1470.259, 'text': "It's not my camera.", 'start': 1467.437, 'duration': 2.822}, {'end': 1472.9, 'text': 'Everything goes through this single point.', 'start': 1471.559, 'duration': 1.341}, {'end': 1476.643, 'text': "This is something what's often called pinhole model or pinhole camera model.", 'start': 1472.94, 'duration': 3.703}], 'summary': 'Pinhole model: all rays from the world pass through a single point.', 'duration': 25.978, 'max_score': 1450.665, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1450665.jpg'}, {'end': 1679.766, 'src': 'heatmap', 'start': 1635.578, 'weight': 1, 'content': [{'end': 1642.843, 'text': 'And this way we can actually go through all the individual parts and, of course, the more things I need to estimate typically,', 'start': 1635.578, 'duration': 7.265}, {'end': 1644.284, 'text': 'the more tricky the things actually can get.', 'start': 1642.843, 'duration': 1.441}, {'end': 1646.505, 'text': 'But all those problems are actually interdependent.', 'start': 1644.744, 'duration': 1.761}, {'end': 1651.849, 'text': 'So if I have one given, it may make a different estimation process easier or more tricky.', 'start': 1646.545, 'duration': 5.304}, {'end': 1656.369, 'text': 'In terms of the cameras, There are two estimates.', 'start': 1652.689, 'duration': 3.68}, {'end': 1658.13, 'text': "One is what's called camera calibration.", 'start': 1656.449, 'duration': 1.681}, {'end': 1665.756, 'text': 'That means we want to estimate the intrinsic parameters of the camera.', 'start': 1658.65, 'duration': 7.106}, {'end': 1669.063, 'text': 'for example, What is the focal length of my camera?', 'start': 1665.756, 'duration': 3.307}, {'end': 1670.443, 'text': "What's the size of a pixel?", 'start': 1669.423, 'duration': 1.02}, {'end': 1673.484, 'text': 'How far is the pixel away from this pinhole?', 'start': 1670.483, 'duration': 3.001}, {'end': 1677.465, 'text': 'And is there any distortion in the mapping, for example from my lens?', 'start': 1674.064, 'duration': 3.401}, {'end': 1679.766, 'text': "That's something which is called camera calibration.", 'start': 1678.005, 'duration': 1.761}], 'summary': 'Estimating intrinsic camera parameters, including focal length and pixel size, is crucial for camera calibration in computer vision.', 'duration': 44.188, 'max_score': 1635.578, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1635578.jpg'}], 'start': 1421.482, 'title': 'Photogrammetry fundamentals', 'summary': 'Covers the pinhole camera model and its role in image analysis, introduces the fundamentals of photogrammetry, and discusses camera calibration and the interdependence of problems in computer vision, all essential for understanding the field. the content emphasizes the physics process of image formation, object localization, and the mapping process, providing a comprehensive foundation for the course.', 'chapters': [{'end': 1536.633, 'start': 1421.482, 'title': 'Pinhole camera model', 'summary': 'Discusses the pinhole camera model, the physics process of getting images, and exploiting background knowledge for image analysis, emphasizing the inverted mapping from images to objects or camera location.', 'duration': 115.151, 'highlights': ['The pinhole camera model is explained, where every ray from the world passes through a single point, resulting in an upside-down image of the scene.', 'The physics process of generating images involves a scene with objects, geometries, a camera with known parameters like focal length and extrinsic parameters, and the process of taking measurements.', 'The discussion emphasizes exploiting physics and background knowledge for image analysis, including the exploitation of principles and additional background knowledge for tasks like classification and semantic analysis.']}, {'end': 1632.994, 'start': 1537.873, 'title': 'Introduction to photogrammetry', 'summary': 'Discusses the tasks and processes involved in photogrammetry, including operations on images, object localization, and reconstruction, which are fundamental in understanding the field of photogrammetry and will be covered in the course.', 'duration': 95.121, 'highlights': ['Object Reconstruction The process of computing the 3D shape and geometry of objects, essential in photogrammetry tasks, and will be covered in the course.', 'Object Localization The task of estimating the position of an object, crucial in photogrammetry, and will be explained in the course.', 'Processing Images The initial step in photogrammetry tasks, involving operations on images, which will be a key focus of the course.']}, {'end': 1764.643, 'start': 1635.578, 'title': 'Camera calibration and estimation', 'summary': 'Discusses camera calibration, estimation of intrinsic parameters, and the interdependence of problems in computer vision, with a focus on understanding the mapping process and the interplay between human perception, eyes, and the brain.', 'duration': 129.065, 'highlights': ['The chapter discusses the process of camera calibration to estimate intrinsic parameters such as focal length, pixel size, distance from the pixel to the pinhole, and distortion in the lens mapping. It explains the need to estimate intrinsic parameters for better mapping accuracy, describing the mapping process from object to image plane.', 'The interdependence of problems in computer vision is highlighted, where having one given estimation may influence the difficulty of other estimation processes. It emphasizes the interconnected nature of estimation problems in computer vision.', 'The chapter explores the interplay between human perception, eyes, and the brain, questioning the division of work between the eye and the brain in processing visual information. It raises the question of whether the eye or the brain does most of the work in processing visual information, prompting an experiment to prove the division of work.']}], 'duration': 343.161, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1421482.jpg', 'highlights': ['The pinhole camera model is explained, where every ray from the world passes through a single point, resulting in an upside-down image of the scene.', 'The chapter discusses the process of camera calibration to estimate intrinsic parameters such as focal length, pixel size, distance from the pixel to the pinhole, and distortion in the lens mapping.', 'The interdependence of problems in computer vision is highlighted, where having one given estimation may influence the difficulty of other estimation processes.']}, {'end': 2110.915, 'segs': [{'end': 1845.075, 'src': 'embed', 'start': 1800.319, 'weight': 0, 'content': [{'end': 1806.042, 'text': 'So the person can actually perceive the scene, understand what they see, what is in front of them,', 'start': 1800.319, 'duration': 5.723}, {'end': 1812.286, 'text': 'just by using their skin as kind of an alternative to using their eyes.', 'start': 1806.042, 'duration': 6.244}, {'end': 1820.81, 'text': 'We can actually show that the person can recognize, for example, different objects or interpret the scene, but with very little training.', 'start': 1813.126, 'duration': 7.684}, {'end': 1830.515, 'text': "And this actually shows that the eyes, although they're very, very good cameras, and the cameras that we have nowadays, at least in most of the cases,", 'start': 1821.951, 'duration': 8.564}, {'end': 1834.137, 'text': 'cannot compete with our eyes, but still the brain does most of the work.', 'start': 1830.515, 'duration': 3.622}, {'end': 1845.075, 'text': 'And this is also one reason why algorithms are so central in photogrammetry because we have cameras, we have images,', 'start': 1835.469, 'duration': 9.606}], 'summary': "Using skin to perceive and interpret scenes, with little training, showing the brain's role in visual processing.", 'duration': 44.756, 'max_score': 1800.319, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1800319.jpg'}, {'end': 1894.042, 'src': 'embed', 'start': 1866.88, 'weight': 4, 'content': [{'end': 1874.345, 'text': "It's a very, very nice, very heavy, very expensive camera that you put in an airplane and you fly over the Earth.", 'start': 1866.88, 'duration': 7.465}, {'end': 1875.606, 'text': 'You take aerial images.', 'start': 1874.585, 'duration': 1.021}, {'end': 1879.094, 'text': 'And this is actually the system which was used to create Bing Maps.', 'start': 1876.493, 'duration': 2.601}, {'end': 1881.095, 'text': "It's exactly those cameras over here.", 'start': 1879.735, 'duration': 1.36}, {'end': 1887.339, 'text': 'And this is kind of the very, very modern version of actually the guy there sitting in the back where I point with the laser pointer.', 'start': 1881.496, 'duration': 5.843}, {'end': 1894.042, 'text': 'So this is the kind of 200 years old, 150 years old version of that, using these large film rolls for taking aerial images.', 'start': 1887.379, 'duration': 6.663}], 'summary': 'High-end camera used for aerial imaging, including in bing maps, with modern and historical versions discussed.', 'duration': 27.162, 'max_score': 1866.88, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1866880.jpg'}, {'end': 1938.209, 'src': 'embed', 'start': 1908.351, 'weight': 5, 'content': [{'end': 1911.393, 'text': 'There are also line sensors which just scan a single line.', 'start': 1908.351, 'duration': 3.042}, {'end': 1920.077, 'text': 'You can actually fly over that and then, by taking stitching line to line, if I know the position of the airplane or the camera,', 'start': 1911.413, 'duration': 8.664}, {'end': 1923.899, 'text': 'I can actually generate those images of the scene.', 'start': 1920.077, 'duration': 3.822}, {'end': 1930.146, 'text': 'If I go to applications on Earth, you have these typical cameras like these ones over here.', 'start': 1925.325, 'duration': 4.821}, {'end': 1938.209, 'text': 'These are just standard cameras approximately this size industrial cameras used for different types of applications, where we can observe objects,', 'start': 1930.166, 'duration': 8.043}], 'summary': 'Line sensors can generate images of the scene by stitching line to line based on the position of the airplane or camera.', 'duration': 29.858, 'max_score': 1908.351, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1908351.jpg'}, {'end': 2090.097, 'src': 'embed', 'start': 2043.629, 'weight': 3, 'content': [{'end': 2047.472, 'text': 'You can actually fly over a scene and use, for example,', 'start': 2043.629, 'duration': 3.843}, {'end': 2058.679, 'text': 'laser scanners or other scanning devices to actually generate a dense 3D reconstruction of the scene and use this for analyzing the terrain using different sensing modalities.', 'start': 2047.472, 'duration': 11.207}, {'end': 2067.797, 'text': 'We can also use, as I said, small flying vehicles like this drone over here to actually monitor land, especially monitor farmland.', 'start': 2060.594, 'duration': 7.203}, {'end': 2072.78, 'text': 'There are quite some applications on this, especially if you look into not only the visible light,', 'start': 2067.818, 'duration': 4.962}, {'end': 2075.862, 'text': 'but also different channels in the infrared spectrum.', 'start': 2072.78, 'duration': 3.082}, {'end': 2080.444, 'text': 'We can actually obtain information about the health of those plants, actually map them.', 'start': 2076.141, 'duration': 4.303}, {'end': 2090.097, 'text': 'generate actually maps of farmland telling me something, what kind of crop is actually growing there and if the crop is healthy or if it is sick.', 'start': 2081.184, 'duration': 8.913}], 'summary': 'Using drones and scanning devices for 3d reconstructions and monitoring farmland health.', 'duration': 46.468, 'max_score': 2043.629, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2043629.jpg'}], 'start': 1765.865, 'title': 'Innovative visual perception', 'summary': 'Covers the process of perceiving scenes without using eyes, emphasizing the significance of algorithms in image interpretation. it also discusses the evolution and modern applications of aerial imaging technology, including generating maps, 3d terrain models, and monitoring farmland.', 'chapters': [{'end': 1866.84, 'start': 1765.865, 'title': 'Seeing with skin', 'summary': 'Discusses a process where a blind person can perceive a scene by using a small resolution camera to project images onto their skin, demonstrating that the brain can interpret visual information without using the eyes, and emphasizing the significance of algorithms in interpreting images in photogrammetry.', 'duration': 100.975, 'highlights': ['The brain can interpret visual information without using the eyes, enabling a blind person to perceive a scene through the use of a small resolution camera projecting images onto their skin.', 'Algorithms play a central role in photogrammetry by interpreting images, emphasizing the importance of developing algorithms to make sense of visual data.', "Blind individuals can recognize and interpret different objects in a scene using their skin as an alternative to using their eyes, showcasing the brain's ability to adapt to alternative sensory input."]}, {'end': 2110.915, 'start': 1866.88, 'title': 'Applications of aerial imaging technology', 'summary': 'Discusses the evolution and modern applications of aerial imaging technology, including the use of high-resolution cameras, laser sensors, and drones to generate maps, 3d terrain models, and monitor farmland.', 'duration': 244.035, 'highlights': ['The chapter discusses the evolution and modern applications of aerial imaging technology, including the use of high-resolution cameras, laser sensors, and drones to generate maps, 3D terrain models, and monitor farmland. The focus is on the evolution and modern applications of aerial imaging technology, including the use of high-resolution cameras, laser sensors, and drones for various purposes.', 'The system used to create Bing Maps involves high-resolution cameras that generate images using different channels, panchromatic as well as color near infrared information. The system used to create Bing Maps involves high-resolution cameras that generate images using different channels, panchromatic as well as color near infrared information.', 'Aerial images are obtained using line sensors which scan a single line and stitch them together to generate images of the scene. Aerial images are obtained using line sensors which scan a single line and stitch them together to generate images of the scene.', 'The technology can be used to monitor land, especially farmland, and obtain information about the health of plants, map crop types, and assess their health. The technology can be used to monitor land, especially farmland, and obtain information about the health of plants, map crop types, and assess their health.', 'The use of small flying vehicles like drones for monitoring land, especially farmland, is discussed, emphasizing the ability to obtain infrared spectrum information about crop health. The use of small flying vehicles like drones for monitoring land, especially farmland, is discussed, emphasizing the ability to obtain infrared spectrum information about crop health.']}], 'duration': 345.05, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY1765865.jpg', 'highlights': ['The brain can interpret visual information without using the eyes, enabling a blind person to perceive a scene through the use of a small resolution camera projecting images onto their skin.', 'Algorithms play a central role in photogrammetry by interpreting images, emphasizing the importance of developing algorithms to make sense of visual data.', "Blind individuals can recognize and interpret different objects in a scene using their skin as an alternative to using their eyes, showcasing the brain's ability to adapt to alternative sensory input.", 'The chapter discusses the evolution and modern applications of aerial imaging technology, including the use of high-resolution cameras, laser sensors, and drones to generate maps, 3D terrain models, and monitor farmland.', 'The system used to create Bing Maps involves high-resolution cameras that generate images using different channels, panchromatic as well as color near infrared information.', 'Aerial images are obtained using line sensors which scan a single line and stitch them together to generate images of the scene.', 'The technology can be used to monitor land, especially farmland, and obtain information about the health of plants, map crop types, and assess their health.', 'The use of small flying vehicles like drones for monitoring land, especially farmland, is discussed, emphasizing the ability to obtain infrared spectrum information about crop health.']}, {'end': 2838.309, 'segs': [{'end': 2196.498, 'src': 'embed', 'start': 2171.564, 'weight': 1, 'content': [{'end': 2178.648, 'text': 'we can actually use the visual odometry information to actually estimate the position of the platform over here.', 'start': 2171.564, 'duration': 7.084}, {'end': 2183.931, 'text': 'So we see that the GPS is lost, you see only the movement generated by the visual odometry.', 'start': 2178.708, 'duration': 5.223}, {'end': 2191.895, 'text': 'And if you then later on reactivate the GPS, you can actually see that they start at very nearby.', 'start': 2184.571, 'duration': 7.324}, {'end': 2196.498, 'text': "so there's a very small drift between the estimate of the cameras and the GPS information.", 'start': 2191.895, 'duration': 4.603}], 'summary': 'Visual odometry used to estimate platform position, minimal drift compared to gps data.', 'duration': 24.934, 'max_score': 2171.564, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2171564.jpg'}, {'end': 2268.216, 'src': 'embed', 'start': 2242.904, 'weight': 2, 'content': [{'end': 2249.867, 'text': "We're extracting features and then doing something which is called bundle adjustment, or in German, which I introduced before,", 'start': 2242.904, 'duration': 6.963}, {'end': 2254.089, 'text': 'when I can actually estimate where are those points in the 3D world.', 'start': 2249.867, 'duration': 4.222}, {'end': 2258.291, 'text': 'We can actually show that those points are actually highly accurate,', 'start': 2254.469, 'duration': 3.822}, {'end': 2268.216, 'text': 'lying actually on the surface of the building and using this for building online maps with those flying platforms and So estimating,', 'start': 2258.291, 'duration': 9.925}], 'summary': 'Extracting features for accurate 3d point estimation and online map building.', 'duration': 25.312, 'max_score': 2242.904, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2242904.jpg'}, {'end': 2384.554, 'src': 'embed', 'start': 2358.191, 'weight': 0, 'content': [{'end': 2364.339, 'text': 'So, estimating at the same point in time the location of all the 3D points and the location of this vehicle,', 'start': 2358.191, 'duration': 6.148}, {'end': 2369.606, 'text': 'you can actually overlay this with maps and actually build highly precise maps using this technique.', 'start': 2364.339, 'duration': 5.267}, {'end': 2377.247, 'text': 'Another example is actually 3D city models.', 'start': 2374.464, 'duration': 2.783}, {'end': 2384.554, 'text': 'so this was quite early work, just driving with cameras and a laser scanner through.', 'start': 2377.247, 'duration': 7.307}], 'summary': 'Estimate 3d points and vehicle location, overlay with maps, build precise maps. early work on 3d city models.', 'duration': 26.363, 'max_score': 2358.191, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2358191.jpg'}, {'end': 2501.044, 'src': 'embed', 'start': 2436.532, 'weight': 3, 'content': [{'end': 2441.434, 'text': "which is equipped here with kind of seven cameras which sit here on the arcs of a lot of light source because it's dark in there.", 'start': 2436.532, 'duration': 4.902}, {'end': 2446.736, 'text': 'So we need to bring our own light sources and can actually drive with this platform through those catacombs.', 'start': 2441.734, 'duration': 5.002}, {'end': 2448.477, 'text': 'Images are a bit dark, sorry for that.', 'start': 2446.976, 'duration': 1.501}, {'end': 2450.058, 'text': 'and actually built.', 'start': 2449.097, 'duration': 0.961}, {'end': 2462.129, 'text': '3D reconstruction uses for digital preservation of cultural heritage and allowing archaeologists actually to observe those models and see for structural problems of those catacombs.', 'start': 2450.058, 'duration': 12.071}, {'end': 2465.498, 'text': 'We can then actually use this to build this kind of 2D maps.', 'start': 2463.137, 'duration': 2.361}, {'end': 2471.961, 'text': 'So this is one of the explore trajectories that we generated in one of those examples.', 'start': 2465.978, 'duration': 5.983}, {'end': 2476.123, 'text': 'Overlay them with, for example, aerial image data.', 'start': 2472.561, 'duration': 3.562}, {'end': 2487.356, 'text': 'Combine this visually obtained data using, for example, three laser scanners, where we get a more accurate 3D reconstruction of those catacombs.', 'start': 2476.883, 'duration': 10.473}, {'end': 2494.286, 'text': 'but this is actually more costly to bring those devices pretty expensive devices, a lot of manual work into those sites.', 'start': 2487.356, 'duration': 6.93}, {'end': 2498.191, 'text': 'These are all different applications that we can actually find here.', 'start': 2494.306, 'duration': 3.885}, {'end': 2501.044, 'text': 'There are also kind of some medical applications.', 'start': 2499.144, 'duration': 1.9}], 'summary': 'A platform with 7 cameras and light sources creates 3d reconstructions for preserving and studying catacombs, while also having medical applications.', 'duration': 64.512, 'max_score': 2436.532, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2436532.jpg'}, {'end': 2547.366, 'src': 'embed', 'start': 2520.668, 'weight': 6, 'content': [{'end': 2526.549, 'text': 'you can actually use x-ray images and analyze those x-ray images to get the right instruments to actually do your surgery.', 'start': 2520.668, 'duration': 5.881}, {'end': 2529.751, 'text': 'You can use all these techniques for quality control.', 'start': 2527.449, 'duration': 2.302}, {'end': 2533.074, 'text': 'For example, you have this arm with high precision measurement devices.', 'start': 2529.991, 'duration': 3.083}, {'end': 2536.637, 'text': 'You can actually move that over a car to do quality control.', 'start': 2533.134, 'duration': 3.503}, {'end': 2539.98, 'text': 'Or other applications for that are robotics applications.', 'start': 2537.738, 'duration': 2.242}, {'end': 2542.182, 'text': 'So like autonomous cars.', 'start': 2540.28, 'duration': 1.902}, {'end': 2543.443, 'text': 'So this is actually the Google car.', 'start': 2542.202, 'duration': 1.241}, {'end': 2547.366, 'text': 'You have automatic lawnmowers.', 'start': 2545.184, 'duration': 2.182}], 'summary': 'X-ray images can be used for surgery instruments, precision measurement devices for quality control, and applications such as robotics and autonomous cars.', 'duration': 26.698, 'max_score': 2520.668, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2520668.jpg'}, {'end': 2670.005, 'src': 'embed', 'start': 2642.099, 'weight': 7, 'content': [{'end': 2645.36, 'text': 'and use this in order to make the best possible decisions.', 'start': 2642.099, 'duration': 3.261}, {'end': 2652.882, 'text': 'And in order to do that, to realize those applications, we really need to know what the environment looks like.', 'start': 2646.26, 'duration': 6.622}, {'end': 2655.263, 'text': 'where are we in the scene, what happens around us?', 'start': 2652.882, 'duration': 2.381}, {'end': 2660.581, 'text': 'And this is just a small demo from the Google car given in 2011.', 'start': 2656.539, 'duration': 4.042}, {'end': 2668.425, 'text': "It's actually a parking lot and the car driving in the parking lot on the top of a parking garage.", 'start': 2660.581, 'duration': 7.844}, {'end': 2670.005, 'text': 'I hope you can hear that.', 'start': 2669.105, 'duration': 0.9}], 'summary': 'In a 2011 demo, a google car navigated a parking lot on top of a parking garage to make informed decisions based on environmental awareness.', 'duration': 27.906, 'max_score': 2642.099, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2642099.jpg'}, {'end': 2782.266, 'src': 'embed', 'start': 2752.728, 'weight': 8, 'content': [{'end': 2758.929, 'text': "you're actually able to understand the overall process and even build parts of these estimation processes on your own.", 'start': 2752.728, 'duration': 6.201}, {'end': 2761.77, 'text': 'And these are kind of key building blocks for a lot of applications.', 'start': 2758.989, 'duration': 2.781}, {'end': 2765.922, 'text': 'There are a couple of relevant literature for this course.', 'start': 2763.541, 'duration': 2.381}, {'end': 2768.662, 'text': 'Just put them on the slides.', 'start': 2767.922, 'duration': 0.74}, {'end': 2771.003, 'text': 'There are actually three things which are pretty important.', 'start': 2768.682, 'duration': 2.321}, {'end': 2774.584, 'text': 'The first one is actually the script by Wolfgang Förstner, which I mentioned before.', 'start': 2771.023, 'duration': 3.561}, {'end': 2782.266, 'text': "Again, this script for photogrammetry one, let's say 90%, is identical to the course material I have.", 'start': 2775.204, 'duration': 7.062}], 'summary': 'Learn to build estimation processes with key building blocks for applications. relevant literature includes script by wolfgang förstner.', 'duration': 29.538, 'max_score': 2752.728, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2752728.jpg'}, {'end': 2818.611, 'src': 'embed', 'start': 2795.071, 'weight': 9, 'content': [{'end': 2803.375, 'text': 'And for the machine learning part, especially these classification approaches, I actually found the Al-Payadin book an excellent technique.', 'start': 2795.071, 'duration': 8.304}, {'end': 2807.217, 'text': "But in terms of the notation, I stick with Wolfgang Foerster's script.", 'start': 2803.415, 'duration': 3.802}, {'end': 2813.48, 'text': "So if you look to this German script, you should find the direct correspondences between what I'm telling here and what happens on the slide.", 'start': 2807.297, 'duration': 6.183}, {'end': 2816.208, 'text': 'So the next lecture is today in a week.', 'start': 2814.687, 'duration': 1.521}, {'end': 2818.611, 'text': "So there's no lecture this Thursday.", 'start': 2816.469, 'duration': 2.142}], 'summary': "Using al-payadin's technique for machine learning classification; next lecture is in a week.", 'duration': 23.54, 'max_score': 2795.071, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2795071.jpg'}], 'start': 2111.535, 'title': 'Applications and techniques in 3d reconstruction', 'summary': "Covers various applications of 3d reconstruction like city mapping and cultural heritage preservation, as well as techniques including visual odometry, bundle adjustment, and autofotos. it also explores the 3d reconstruction of santa priscilla catacomb, involving the use of cameras, light sources, and visual data for digital preservation and archaeological observation. additionally, it delves into the google car's environment interpretation techniques and their applications in decision-making scenarios, referencing relevant literature.", 'chapters': [{'end': 2420.222, 'start': 2111.535, 'title': 'Applications of 3d reconstruction', 'summary': 'Discusses the use of 3d reconstruction techniques, such as visual odometry, bundle adjustment, and autofotos, for mapping and surveying purposes, including city mapping and preservation of cultural heritage.', 'duration': 308.687, 'highlights': ['Visual odometry for estimating platform position Visual odometry is used to estimate the position of the platform, providing a backup when GPS fails and showing minimal drift compared to GPS estimates.', 'Use of bundle adjustment for 3D point estimation Bundle adjustment technique is applied to estimate 3D points in the environment, enabling accurate alignment on surfaces and construction of online maps with flying platforms.', 'City mapping and surveying with 3D reconstruction 3D reconstruction techniques are utilized for city mapping, surveying, and building 3D city models at various scales, demonstrating the practical application of 3D reconstruction in mapping larger areas.', 'Applications in preservation of cultural heritage The use of 3D reconstruction techniques is mentioned in relation to the preservation of cultural heritage, indicating diverse applications beyond mapping and surveying.']}, {'end': 2542.182, 'start': 2420.222, 'title': '3d reconstruction of santa priscilla catacomb', 'summary': "Discusses the use of a platform equipped with cameras and light sources to navigate and create 3d reconstructions of the dark catacombs, aiding in digital preservation and archaeologists' observation. it also explores the use of visual data, laser scanners, and various applications such as medical, quality control, and robotics.", 'duration': 121.96, 'highlights': ["3D reconstruction of catacombs using a platform with cameras and light sources The platform equipped with seven cameras and light sources navigates through catacombs to create 3D reconstructions, aiding in digital preservation and archaeologists' observation.", 'Use of visual data and laser scanners for more accurate 3D reconstruction Visual data and laser scanners are utilized to overlay explore trajectories and obtain a more accurate 3D reconstruction of the catacombs.', 'Applications in medical field such as dentistry and surgery The techniques are applicable in analyzing x-ray images for root channel estimation in dentistry and surgery, aiding in dental and medical procedures.', 'Quality control and robotics applications The techniques can be employed for quality control, such as using high precision measurement devices for automotive quality control, and in robotics applications like autonomous cars.']}, {'end': 2838.309, 'start': 2542.202, 'title': 'Google car and environment interpretation', 'summary': "Discusses the google car's use of techniques for environment interpretation, including estimating objects and interpreting data using cameras and classification approaches, to make decisions in different scenarios. it also introduces the goal of the course to teach these techniques and references relevant literature.", 'duration': 296.107, 'highlights': ['The Google car uses techniques for environment interpretation, including estimating objects and interpreting data using cameras and classification approaches. The prototype of the Samsung Hudson Hoover uses small cameras to estimate objects on the ceiling and localize to find its way back to the charging station, demonstrating the need for environment interpretation techniques.', 'The goal of the course is to teach techniques for environment interpretation and introduce students to key building blocks for various applications. The goal is to introduce students to different techniques for understanding the overall process of environment interpretation and building parts of the estimation processes, which are key building blocks for a lot of applications.', "Reference to relevant literature, including scripts by Wolfgang Förstner and the Al-Payadin book for machine learning approaches. The course references Wolfgang Förstner's script for photogrammetry and the Zelinsky book on computer vision, along with the Al-Payadin book for machine learning approaches in terms of classification."]}], 'duration': 726.774, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/_mOG_lpPnpY/pics/_mOG_lpPnpY2111535.jpg', 'highlights': ['3D reconstruction techniques for city mapping and surveying', 'Visual odometry for estimating platform position as a backup to GPS', 'Bundle adjustment technique for accurate 3D point estimation', 'Applications of 3D reconstruction in preservation of cultural heritage', '3D reconstruction of catacombs using cameras and light sources', 'Visual data and laser scanners for accurate 3D reconstruction', 'Applications of techniques in dentistry, surgery, quality control, and robotics', "Google car's environment interpretation techniques and decision-making applications", 'Teaching environment interpretation techniques and key building blocks', 'Reference to relevant literature including scripts by Wolfgang Förstner and Al-Payadin book']}], 'highlights': ['Photogrammetry involves estimating geometric and semantic properties of objects based on images or similar sensors.', 'The importance of capturing multiple images with camera sensors and exploiting them to estimate 3D geometry and changes in scenes, such as in smartphones recording videos at 25-30 frames per second.', 'Contact-free sensing allows for large area coverage, making data acquisition very easy.', 'The pinhole camera model is explained, where every ray from the world passes through a single point, resulting in an upside-down image of the scene.', 'The brain can interpret visual information without using the eyes, enabling a blind person to perceive a scene through the use of a small resolution camera projecting images onto their skin.', '3D reconstruction techniques for city mapping and surveying', 'Visual odometry for estimating platform position as a backup to GPS', 'Bundle adjustment technique for accurate 3D point estimation', 'Applications of 3D reconstruction in preservation of cultural heritage', 'Applications of techniques in dentistry, surgery, quality control, and robotics']}