Research On Detection Method Of Surface Defects Of Key Parts Of Office Chair Based On Deep Learning

In recent years,the scale of development of the domestic office chair industry is expanding,in the actual production process office chair key components usually produce different types of defects.At present,the detection and identification of office chair key components surface defects are mainly through the manual naked eye approach,low production efficiency;some traditional digital image processing detection methods often occur leakage and false detection and other problems,detection accuracy needs to be improved.Therefore,improving the efficiency and accuracy of detection and recognition is an urgent problem to be solved.In the rapid development and continuous progress of computer and artificial intelligence,the application of deep learning-based target detection methods in the field of manufacturing has obtained good research results.In this regard,this paper proposes a deep learning-based surface defect detection method for key components of office chairs,with the following main research contents.（1）Analyze and compare the performance of multiple types of target detection algorithms through publicly available data sets.For this topic in the office chair key parts surface defects of more types and more complex backgrounds and other characteristics,the algorithm selected for this paper is the YOLOv3 algorithm.（2）To collect the image information in the field,the original images were subjected to uneven sample size and other problems due to the actual production conditions in the factory,and the original images were enhanced to expand the data set,including mirroring,panning,rotation,dimming,cropping,noise addition and grayscale.The data is labeled by Label Img software and divided into8:1:1 ratios to obtain the standard dataset.（3）Construct the network structure model of YOLOv3,feed the dataset into the network model,train the YOLOv3 network model and deeply analyze the research results,and then form the optimization scheme of YOLOv3 by combining the actual situation of surface defects of key parts of office chairs,mainly including the modification of the network structure of the multi-scale prediction part,proposing the use of k-means||algorithm to re-cluster the dataset to obtain suitable anchor boxes,propose to use CIOU to replace the original IOU loss function,and propose Merge DIOU-NMS to replace the original NMS function.After experimental validation,the proposed Improved YOLOv3algorithm achieves an m AP value of 94.85%and a detection speed of 47.62 f·s^-1 on the test set in the foot of key office chair components.to further verify that the proposed algorithm can adapt to multiple types of key office chair components,further validation was done on the bottom shell dataset and the five-star foot dataset,and the m AP values respectively The results show that the Improved YOLOv3algorithm can effectively detect the surface defects of key office chair components.（4）Complete the construction and application of the inspection platform.Select suitable lens,industrial camera,light source and lighting method,build hardware platform,adopt Python language and design and develop software system based on Py Qt5 to realize the image acquisition and defect detection recognition of surface defects of key office chair parts.