为实现导线端子 － 连接器插孔装配的自动化，设计了基于机器视觉的机器人自 动插线系统。为克服成像视角变化及杂乱背景对连接器目标识别造成的干扰，提出一种 基于累积量化梯度方向特征的连接器识别方法，提高了目标连接器识别的稳定性; 在此 基础上，针对因插孔密集排列造成的多插孔中心定位问题，提出一种基于扫描线直方图 特征的多插孔定位算法，实现了目标连接器插孔中心位置的精确获取。实验结果表明， 所提出的算法能够稳定识别目标连接器，并满足插孔定位的精度要求。

In order to realize automated wire-socket assembly，an automatic robot wiring system based on machine vision was designed． And，to avoid the interference caused by the change of imaging angle and the clutter background，a connector identification method based on cumulative quantization gradient orient feature was proposed to improve the stability of target connector identification． On this basis，a multi-socket positioning algorithm based on scan line histogram characteristics was proposed to solve the problem of multi-sockets' center positioning caused by the dense arrangement of sockets． Experimental results show that the designed algorithm can stably identify the target connector and meet the accuracy requirements of socket positioning．