Detection of Moving Objects by Spatio-Temporal Motion Analysis

Driver assistance systems of the future require a thorough understanding of the car's environment. In this thesis, a novel principle (6D-Vision) is presented and investigated in detail, which allows the reconstruction of the 3D motion field from the image sequence obtained by a stereo camera system. Given correspondences of stereo measurements over time, this principle estimates the 3D position and the 3D motion vector of selected points using Kalman Filters, resulting in a real-time estimation of the observed motion field. To estimate the absolute motion field, the ego-motion of the moving observer must be known precisely. Thus, a novel algorithm to estimate the ego-motion from the image sequence is presented. As the 6D-Vision principle is not restricted to particular image processing algorithms, various optical flow and stereo algorithms are evaluated. In addition, two novel scene flow algorithms are introduced, measuring the optical flow and stereo information in a combined approach. This yields more precise and robust results. The application to real-world data, including a demonstrator vehicle for autonomous collision avoidance, is illustrated throughout the thesis.