基于图像识别的微小塑料齿轮检测研究

Research on the On-line Inspection of Micro Plastic Gear Based on Computer Vision

采用图像识别技术对微小塑料齿轮进行质量检测,针对缺齿、齿歪、披峰等齿形误差的随机性,从计算机视觉系统的构成、数字图像采集、图像预处理和图像分割理论与技术等方面进行深入研究,提出迭代阈值与最大类间方差区域切割综合法,解决了双联齿轮的图像分割难题,其算法适应在线实时检测的速度要求。提出虚拟圆扫描法,实现双联齿轮齿形的缺陷检测。结果表明,该检测技术的缺陷识别率达95%。

The quality of micro plastic gears was inspected with the image recognition technology herein. Focusing on the situation that gears＇ defects were uncertain, the construction of computer vision system and the theories and technologies of digital image acquisition and image preprocessing as well as image segmentation were studied thoroughly. Colligatlng the iterative threshold and Otsu arithmetic, the difficult problem about the image segmentation of duplicate gear was solved, which can meet requirements on the speed for real-time inspection of the micro plastic gears. This paper also presented a method, dummy circle scan method, to realize the inspection of duplicate gears＇ defects. Experimental result shows that the identification rate of defects of the system reaches as high as 95%.