Research and Application for Correctness Inspection of Control Cabin Assembly based on YOLOv5
Aimed at the problems that the assembly process of a certain type of control cabin mainly adopted manual in-spection,the detection results were affected by human factors,which could not completely guarantee the assembly quality of the product,and the level of detection automation was not high,it was proposed that an on-line detection system to im-prove the correctness of the assembly of the control cabin of YOLOv5.This system consisted of two parts,which were hardware and software.The hardware included server,industrial computer,CCD camera,buzzer and so on,and the soft-ware consisted of detection module,Lab VIEW control program and MES system.For the problem of noise points in the im-age,we added the CBAM attention mechanism module on the basis of the original YOLOv5 detection algorithm,which made the network able to locate and recognize the region of interest more accurately,so as to improve the detection accuracy of the detection model.The experimental results showed that the improved YOLOv5 detection algorithm significantly im-proved the accuracy of the online detection of the correctness of the control pod assembly,comparing with the original algo-rithm,the improved YOLOv5 algorithm achieved 99.1％in the detection accuracy,which was better than other algo-rithms,and it was 2.4,0.9,0.5,and 0.3 percentage points higher than SSD,YOLOv3,YOLOv4,and YOLOv5,respec-tively.The average accuracy mAP reached to 99.4％that was improved by 0.4 percentage over the original algorithm.The proposed improved YOLOv5 control module assembly correctness online detection algorithm was able to detect the correct-ness of parts assembly in the process of control module assembly in a timely manner,which provided technical support for the wide application of deep learning technology in the control module assembly correctness online detection.
万方数据
维普
