面向机器人航天员的3D目标识别定位算法参数优化研究

摘要

This paper aims at the fast recognition and location issue of the robot astronaut when grab⁃bing objects. The main applications of the 3D target recognition and location algorithm in HALCON based on CAD models used in the vision of American Robonaut 2 were described. Then the parame⁃ter optimization of the algorithm was studied. The effect of various parameters on the algorithm was analyzed. In addition, the selection principle and harmonious collocation of all parameters were dis⁃cussed. Finally, an experimental platform was set up and it was validated that the proposed optimi⁃zation improved the similarity of template matching, achieved 100% correct recognition rate, and only spent more than 200 ms for the megapixel image. The mean position accuracy was less than 2 mm and the orientation accuracy was less than 1 degree.%针对机器人航天员抓取目标物体时的快速识别定位问题，分析了 HALCON 中基于CAD模型的三维目标识别与定位算法在美国Robonaut 2视觉上的应用；对该算法的参数优化进行了研究，分析了各个参数对算法的影响效果，并对各个参数的选择原则以及协调搭配进行了探讨；最后搭建实验平台，验证了优化效果，证明这些优化提高了模板匹配的相似度，达到100％正确识别率，处理百万像素图像仅需约200 ms，且位置平均精度小于2 mm，角度平均精度小于1°。