Research On Battery Internal Defect Detection Technology Based On Image Analysis

As the application scale of battery is becoming wider and larger,higher requirements are placed on the battery safety assessment.As an important means to realize non-destructive testing of batteries,image analysis technology is widely used in the detection of internal defects of rechargeable batteries in the domestic and overseas areas,which can effectively improve the detection speed and accuracy.However,the research on disposable batteries is still in the early stage,such as thermal batteries.Therefore,the X-ray images inside the thermal batteries were taken as the detection object,the identification and classification methods of internal defects of the thermal battery were studied and a software system for detecting defects of the thermal battery was developed in this paper.The main research contents are as follows:Thermal battery defects are mainly present in internal monomer battery,and each of the individual monomer battery needs to be accurately and completely separated.When dividing the monomer thermal battery,the X-ray image of the thermal battery was pre-processed,and the corners of the stack were preliminarily extracted by the template matching method.Then,the cell stack was adjusted by the algorithm of corner detection and rotation to ensure that the background portions were filtered out from all the images of every single thermal cell.In view of the characteristics that the single cell is relatively thin,a top hat algorithm and iterative binarization were used to highlight the cell boundary.Finally,each cell image was extracted by an improved scanning method.The experimental results showed that this method is suitable for monomer heat.The segmentation accuracy of the battery reached97.6%,and the positioning accuracy reached 1.01 pixel on average,which satisfies the requirements of battery detection and segmentation.Three assembly defect characteristics of single-cell thermal battery were studied In this paper.The Hu invariant moment,improved GLCM and template matching algorithm were selected to extract the shape characteristics,the texture features and the position parameters of the defective battery,which realized the detection of this three assembly defects.The parameters were determined by the weighted average method based on the characteristic parameters of each algorithm.The experimental results showed that the accuracy of this method was 97.5%,which had higher accuracy and wider adaptability than the single feature detection method.This method met the testing requirements of single thermal battery,and provided an effective way for thermal battery defect detection.For the detection and identification of thermal battery defects,two types of classifiers were used in this paper: BP neural network and CART(classification and regression tree)decision tree,and in-depth study of classifier hidden layer and training function was carried out.The parameters of each classifier were set by using theoretical analysis and experimental studies.At the same time,in order to get better classification effect,the preferred method was selected as the classifier according to the classification accuracy of each classifier for different detection methods of single thermal battery,and the accuracy rates of the final Hu invariant moment,improved GLCM and template matching method reached 89.33%,95.50%and 95.00%,respectively.At the end of this paper,the image processing theory was directly combined with the modules in the software system,and a simplified interactive interface design was carried out.The detection technology of internal defects of thermal batteries was studied in this paper based on the image analysis methods.The single-cell thermal battery was completely segmented by background segmentation and improved scanning method.The effects of three feature extraction schemes on battery classification were analyzed.The accuracy of the three detection methods was improved by the classifier optimization method.The technology made the accuracy of the single-cell thermal battery classification detection finally reach 97.5%.The detection scheme provides an effective way for the classification and identification of internal defects of the thermal battery.