Robot Vision

The course covers a wide array of topics starting from human visual perception, camera mechanics, optics and illumination, and the basics of image quality through sampling, quantization, and standards. It delves into digital image processing for image restoration, employing techniques like smoothing, sharpening, filtering, and transformations. The course addresses robust object recognition, based on keypoints, edge and shape detectors, and deep learning neural networks. It explores 3D model alignment and (self-)calibration of imaging systems. Further, it introduces 3D object reconstruction using stereo vision, structured light and photometric stereo. Visual navigation is discussed through image-based tracking, motion analysis, pose estimation, self-localization and mapping. Finally, it delves into image-based decision making via semantic segmentation, feature extraction, and machine learning based classification, highlighting applications in visual quality control.

The objective of this course is to: learn about technologies, devices, and processes for visual perception with modern imaging techniques, digital image processing, understanding of image content, robotic measurement and calibration, object recognition and tracking, navigation, and visual quality control. Students will gain confidence in independently designing, constructing, and critically evaluating industrial robotic vision systems and solving related problems through theoretical foundations and solving practical examples. 

Theoretical foundations, concepts and exemplar use cases are given during lectures, while practical skills are gained through lab works, weekly labwork assignments and an individual seminar work.