Obstacle detection based on single view vision
Liu Ji-lin · Journal of Zhejiang University(Engineering Science) · 2008
By extending the factorization methods for the structure from motion(SFM) to the obstacle detection in outdoor scenes,a novel obstacle detection method based on a single camera was presented.After the feature points in the image sequence were extracted and tracked,the projective structure was obtained from the factorization-based SFM estimation algorithm.Then the self-calibration based on dual absolute quadric(DAQ) was used to determine the internal parameters of the camera in order to recover the camera motion and upgrade the projective structure to the metric structure.Finally,the image was segmented into regions with the same size.Each individual region was classified as belonging to either an obstacle or the background based on its depth information gained from Euclidean reconstruction.Experimental results in real outdoor scene show that this method can effectively detect obstacles.