Research On Pin Detection Technology Of Automotive Electronic Connectors Based On Machine Vision

Automotive electronic connectors are widely used in automotive power and signal systems,and play a role in circuit control and signal transmission.The reliability of the quality of electronic connectors involves the safety of the entire vehicle and is an extremely important part of the automotive circuit.Strict quality testing is required for connectors before assembly.In view of the fact that the quality inspection of traditional automotive electronic connectors mainly relies on simple manual visual inspection,the inspection of connectors has problems of high labor cost,low efficiency,inconsistent inspection standards,and low inspection accuracy.Therefore,this paper proposes an automatic connector pin detection method based on the combination of machine vision and deep learning to achieve fast,high-precision and automated detection of connectors.This thesis studies the pin identification and pin positioning of electronic connectors.The main research contents of the paper are as follows:(1)Aiming at the problem of high-quality image acquisition in the automated visual inspection system of electronic connectors,the image acquisition scheme and lighting scheme of the machine vision inspection system were studied in detail.Visual inspection platform.(2)Aiming at the problem of small pin targets in the collected connector images and the difficulty of pin detection and identification by traditional image processing methods,a pin identification algorithm based on improved YOLOv3 is proposed.By introducing the Dense Net network structure in the YOLOv3 backbone network,it optimizes the transfer of feature information,enhances the learning ability of the model,and relieves the phenomenon of gradient disappearance;at the same time,it optimizes the detection of small targets such as connector pins and adjusts the prediction scale of the model,To improve the accuracy of the pin recognition model.Experimental results show that the improved YOLOv3-Dense network model has higher detection accuracy and detection speed.(3)Aiming at the problem of low accuracy of pin positioning using traditional methods for connector images,a pin sub-pixel positioning method based on improved Zernike moment and mean shift is proposed.First,the pin target features are analyzed to realize the background separation of the pin image and the pin interest area is extracted;secondly,the improved Zernike moment method is used to extract the sub-pixel edges of the pin,combined with Mean Shift clustering algorithm and least square Sub-pixel positioning of pin locations.The experimental results show that the positioning accuracy of the method to the connector pins is within 0.1 pixel.