tailieunhanh - Advances in Theory and Applications of Stereo Vision Part 2

Tham khảo tài liệu 'advances in theory and applications of stereo vision part 2', kỹ thuật - công nghệ, cơ khí - chế tạo máy phục vụ nhu cầu học tập, nghiên cứu và làm việc hiệu quả | Evolutionary Approach to Epipolar Geometry Estimation 15 epipolar geometry is exploited to measure the motion of a camera in the three dimensional space the solution search space of the genetic algorithm can be dramatically reduced. If the camera at the preceding time step was heading in a given direction with a given pose at the following one will head toward a direction and with a pose not much dissimilar from the preceding ones with a very high probability. This means that the initial population of roto-translation can be tuned on the basis of the near motion history of the camera and the algorithm will quickly converge to the right solution. With a standard approach the motion information would have been of no use since the epipolar geometry estimation would have been performed blindly processing all the data points. In different words using the present approach it is possible to add constraints on the geometry of the camera system. The genetic design of the algorithm is responsible for the capability to automatically ignore outliers during the computation. The tness computation is based on the geometric distance between points and epipolar lines relative to their correspondences. If a given roto-translation is the one accurately describing the geometry then all the correct correspondences will have a quasi null distance and the outliers will yield a large one. This mechanism can be exploited to automatically pin out the outliers from the set of available experimental correspondences also implying the possibility to reduce the data set to the inliers only gaining in precision and on the quantity of data to process. The results in terms of performance are good. In Section 5 it has been shown that the error is not larger than that of the best performing algorithm presently available in literature that the robustness against noise in the data is good and also the insensitivity against the presence of outliers among the data point is high naturally. The algorithm .