基于立体视觉的空间球体快速定位方法

Fast Spherical 3D Location Based on Stereo Vision

采用双目立体视觉系统,对空间球体三维视觉定位方法进行了研究,提出了基于双视锥曲面的球体定位方法.为了实现对球体三维位置的动态实时检测,提出一种基于预测的椭圆曲线快速搜索算法.该算法采用随机Hough变换的取样思想和启发函数预测椭圆搜索区域,减少Hough变换的计算量.为了提高Hough表的操作效率、减少存储空间,设计了一种有效的Hough表存储结构.通过多种方法的综合应用,降低了球体三维定位的计算量和对存储空间的需求,并在实验中验证了所提出的方法.

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.