Energy Meter Patch Resistance and Welding Spot Anomaly Detection Method Based on Machine Vision

Abstract Based on the problems, such as the difficulty of detecting and obtaining evidence of patch resistance replacement and welding spot anomaly in the current field of meter anomaly detection, a patch resistance and welding spot anomaly detection algorithm is proposed based on machine vision. The resistance anomaly detection algorithm combines the K-D tree and Ransac to complete the high-efficiency energy meter registration. It detects the suspected resistance abnormal area through the difference shadow method and then judges the resistance abnormal situation according to the resistance value recognized by the classification network. The welding spot anomaly detection algorithm enhances the feature of the image through saliency detection, then obtains the target information of the welding spot by segmentation, and finally determines the welding spot anomaly condition in combination with the connection domain analysis. Experimental results show that the precision of patch resistance anomaly of this method reaches 95.28%, and the detection time is about 1.52s; the precision of welding spot anomaly reaches 96.74%, and the detection time is about 0.74s. The method can meet the requirements of spot detection accuracy and speed and has good application value.