The paper aims to design a detection system for harness end coating quality based on machine vision on the basis of industrial robot and industrial camera to automatically and accurately identify the end coating quality of au-tomotive harnesses. The CCD industrial camera placed in the industrial robot arm was used in the system to collect image information. First, the system was calibrated by HALCON. Secondly, the gradient coordinates and other preprocessing, corner detection and other methods were used to obtain the node coordinates of the harness. And then, based on the normalized matching standard, the multi-region double-layer multi-threshold template matching algorithm was proposed to identity the end-wrapping quality of the harness quickly, calculate the uncoated length of the end of the harness, and obtain the correction value of the harness node. Finally, the unwrapped segment was replenished. The experiment showed that the accuracy of the coating quality at the end of the harness with less than 10 nodes at the end of the detection system was maintained above 97.22%. The absolute value of the deviation after the replenishment was less than 2 mm, and the repeatability accuracy was 1.07 mm. The detection system is stable and reliable, can meet the detection requirement of wire bundle coating quality in actual production. It improves the automation and intelligence level of the harness coating production line.