Research Of Welding Spot Defection Detection Based On Neural Network

With the continuous progress of science and technology,all kinds of electronic products emerge in an endless stream,and the quality requirements for electronic products are getting higher and higher.Electronic products in the welding process will appear a variety of defects,so it is essential to spot defect detection.Traditional detection methods are mostly based on manual detection,with low efficiency and poor accuracy.The method of defect detection combined with deep learning technology and machine vision can make up for these shortcomings.In this paper,a defect detection algorithm based on convolutional neural network is proposed,which firstly locates the defect parts roughly,determines the defect location information,and then performs precise semantic segmentation of the defect parts.Algorithm in this paper by feature extraction,regional advice network and semantic network of three parts,the segmentation algorithm of overall process is first through the network for feature extraction of feature extraction,and then suggest that network used for coarse positioning faulty areas,get the position information of defective parts,the location of the defect parts information transmission to the semantic network segmentation,carries on the precise pixel level division.In this paper,welding spot defects are defined,the original image is marked with labelme tool,and in order to better train the network,the data set enhancement method is used to expand the data set.In this paper,the operation speed of the algorithm is optimized.The convolution calculation process is replaced by the combination of channel separation convolution and single point convolution,the up-sampling operation and channel convolution combination,which reduces the computation amount of the network to some extent and reduces the detection time of the network.In this paper,the algorithm is tested on the self-built data set,and its accuracy is 98.4%.,which proved that the proposed algorithm is effective for spot defect detection.