稳定杆自动化生产的视觉识别定位研究-中国科技论文在线.pdfVIP

中国科技论文在线 稳定杆自动化生产的视觉识别定位研究# 赵韩，戴智亚，甄圣超，孙国强，周长斌** （合肥工业大学机械工程学院，合肥 230009 ） 5 摘要：为了提高汽车横向稳定杆的生产效率，提出了一种基于单目立体视觉的汽车横向稳定 杆生产线的改造方案，并对该方案的原理、被测零件的识别与位姿求解进行了研究。根据单 目立体视觉原理，对被测零件进行间隔拍摄，利用本文提出的基于背景差分法的识别方法对 被测零件进行识别，并通过识别结果判断是否可以进行位姿计算。利用本文提出的基于图像 质心偏移以及基于图像边缘偏移的方法计算被测零件的位姿信息，再结合摄像机标定结果得 10 到被测零件的实际位姿。实验结果显示，共选取计算出7 个连续时刻的零件位姿，整个识别 定位过程共耗时0.874s ，满足实时测量的要求。 关键词：单目立体视觉；稳定杆；背景差分法；位姿识别；摄像机标定 中图分类号：TP391.41 Research on the visual recognition and location of the 15 automatic production of the stabilizer bar ZHAO Han, DAI Zhiya, ZHEN Shengchao, SUN Guoqiang, ZHOU Changbin (School of Mechanical Engineering, Hefei University of Technology, Hefei 230009) Abstract: To improve the production efficiency of the stabilizer bar, a transformation program of the production line of stabilizer bar based on single camera stereo vision was proposed, and its principle, the 20 recognition and pose solution of the tested parts were studied. According to the principle of the single stereo vision, the parts were measured for interval. Then the recognition method based on background subtraction is proposed to identify the tested parts, and determine whether the pose can be calculated through the results of recognition. Using the proposed method based on the centroid offset and the edge shift of the image to calculate the pose of the parts and get the practical pose of the parts combined with 25 camera calibration results. The results of the experiment showed that seven continuous poses of the parts were selected and calculated. The whole process took a total of 0.874 second and met the measuring requirement in real time. Keywords