基于双目视觉的多型号电连接器检测技术

通过双目视觉平台捕获图像反馈PC机，采用基于支持向量机（SVM）的机器学习方法对目标进行分类。判定型号并为每个型号加载图像处理检测方案及参数矩阵，实现电连接器针脚的柔性定位预处理，之后执行针顶轮廓的精确拟合算法，结合插针排列模版实现其三项检测工作。最后，通过实例验证与重复精度实验，结果表明本文方法具备面向多型号的柔性检测能力，并且稳定性，精度，效率满足低频连接器的检验要求。

Research on Multi-type Electrical Connectors Detection Based on Binocular Vision

Capturing images via binocular vision hardware and transfering data to PC, the system implement connectors classiifcation based on support vector machine (SVM). The connector has been identiifed. The system load image processing algorithm and parameter matrix to locate the pins approximately. After performing precise pins contour iftting algorithm, the system achieve its three main function via comparing the detected data with template. The examples of veriifcation and repeatability of experiments show that the method described above have the ability of multi-type electrical connectors detection and meet the spacecraft's low frequency connector inspection requirements from stability, accuracy, efifciency.