Research On The Defect Detection Method Of Lithium Battery Protection Plate Nickel Sheet Based On Machine Vision

Mobile phone lithium battery protection board(PCB)is an important component to ensure the safety of mobile phones.In the production process,due to the poor incoming precision and consistency of the lithium battery protection board,the defects of positive and negative nickel sheets on the PCB will affect the assembly accuracy of the subsequent welding,making it difficult to put the nickel sheets into the welding vehicle after clamping,resulting in the decline of product yield.At present,most domestic protective board manufacturers are still using manual inspection to detect defects.Therefore,based on the visual inspection requirements of nickel sheet defects in PCB production line,this paper designs a hardware system for nickel sheet defect detection according to the production environment,and studies a nickel sheet subpixel edge extraction method based on Canny operator and bilinear interpolation.And by comparing the area features and SVM to classify the nickel sheet,the software platform of nickel sheet detection system was built.The main work is as follows:Firstly,the overall scheme of nickel sheet defect detection of lithium battery protection plate is designed.By analyzing the assembly requirements of nickel sheet production process,four types of nickel sheet states were divided: normal nickel sheet,nickel sheet with an Angle greater than 93°,nickel sheet with an Angle less than 87°,and folded nickel sheet.The latter three types were defined as defects.In this paper,the size of nickel chip is measured by pixel,and the overall scheme of nickel chip detection is designed.By simulating the shooting state of the second joint of the mechanical arm of the production line fixed camera,the camera and lens were selected,and the appropriate light source and illumination mode were selected.Secondly,the subpixel edge extraction method of nickel slice is studied.The processing effects of different filters and threshold segmentation algorithms are compared.The edge extraction effects of Canny operator,Roberts operator,Sobel operator and LOG operator are compared.The sub-pixel edge is further extracted according to bilinear interpolation method.The experimental results show that under the same nickel slice image,the sub-pixel edge precision extracted by Gaussian filtering method,maximum inter-class variance method,Canny operator and bilinear interpolation method can reach 0.1pixel,which can meet the demand of nickel slice edge precision.Thirdly,the nickel sheet classification method is explored based on area features and SVM.Area extraction algorithm is used to extract the area features of the subpixel edge of the nickel slice,and two types of nickel slice Angle are divided: too large and too small.Aiming at the normal and folded states that the area features could not distinguish,the texture features were extracted as the input of the classifier,the support vector machine model was designed,and the radial basis kernel function was introduced for classification.The experimental results showed that the classification accuracy based on SVM reached 98%,and the calculated Kappa coefficient was 0.9607,indicating that the model had high consistency.Finally,the detection system software platform of nickel sheet is built.Halcon and Visual studio software were combined to design the algorithm interface and detection interface.The algorithm interface displays common algorithms for threshold segmentation,edge extraction,and filtering,and the detection interface displays real-time processing results and displays the area,length,width,processing time,and defect type parameters of the nickel sheet.