Research Of Motion Obstacle Detection Algorithm Under The Dynamic Scene Of Monocular Vision

As cars become a part of our life, traffic accidents become a world-wide problem. Obstacle Detection Systems turn into a world-wide important area as systems solving traffic safety problems. Movement obstacle becomes a hot problem as its importance in the obstacle detection systems. By summing up a large of research methods at home and abroad, a new type of movement obstacle detection algorithm is presented in this thesis.The algorithm includes two parts:the global-motion estimation part and the motion obstacle detection part. The whole idea of the algorithm is:first, the global motion is estimated caused by camera motion under the dynamic scene. Then, the estimation is used for the motion obstacle detection part and realizes the final aim of detecting obstacle.In the global-motion estimation phase, Firstly, based on SURF algorithm, the space coordinates and the matching relations between the feature points are obtained. Then use double linear model and the least-square method, calculating the parameters model of global motion. And mismatching problem encountered in the process of practical application, adaptive to remove mismatching solutions is put forward. Finally, through to the real image experiments, results show that the thinking of global motion estimation is combined with SURF algorithm has the rationality, this remove mismatching method putted forward in this thesis is effective and feasible, and has obtained more satisfactory results.In the motion obstacle detection phase, a motion obstacle detecting method under the dynamic scene is presented based on Bayes decision. Firstly, Bayes decision is used to set up a background detection probability model based on gray level characteristics. Then, according to the result of the global-motion estimation foreground candidate could be judged as the movement information. Finally, combining with the gray level characteristics and movement information, this thesis proposes a novel feature statistics update method, complete movement obstacle detection under the condition of no more constraint conditions. Experimental results show that the algorithm is not easily influenced by illumination conditions, and it not only could improve motion obstacle detection rate, but also effectively reduce false detection rate.