Fast Spherical 3D Location Based on Stereo Vision

Spherical 3D location with binocular stereo vision system is adopted.An algorithm of spherical 3D location based on two view cones is proposed.Basic principles and realization of the method are introduced.In order to realize dynamic real time detection of spherical 3D location,a fast searching algorithm for elliptical curve based on prediction is proposed.The sampling approach of the algorithm is the same as those in random Hough transform where a heuristic function is employed to predict the searching region in the image,much lessening the computing load.To further improve the efficiency of Hough transform table and reduce the memory usage,a new table structure for storing the transformation data is proposed.The computing load and memory usage are greatly reduced by combining these methods.Experiments confirmed the feasibility of the proposed method.