Research On Detection Algorithm For Surface Defects Of Lithium-ion Battery Electrode Based On Machine Vision

In recent years,with the rapid development of portable electronic equipment industry and the rise of new energy technology and new energy vehicle industry,the demand for lithium batteries has increased significantly.As the core part of lithium battery,the surface defects of lithium battery electrode will affect the performance and lifetime of lithium battery in different degrees,what is more,it will bring security risks.Therefore,it is crucial to detect the surface defects of lithium battery electrode in the actual industrial production process.However,the traditional manual detection method is time-consuming and inefficient,and it is also extremely prone to defect missed detection and false detection.Therefore,this paper studies the detection algorithm for surface defects of lithium battery electrode and proposes a machine vision-based automatic online inspection scheme for surface defects of lithium battery electrode to replace manual detection.The defect inspection scheme proposed in this paper is composed of hardware platform and software algorithm.The hardware platform part is principally the design and construction of the hardware platform composed of the image acquisition system,the electrode transmission system and the control system,as well as the selection of the corresponding equipment in the image acquisition system.The software algorithm part is mainly composed of three parts: image preprocessing,defect detection and defect classification.(1)The image preprocessing part firstly extracts the ROI region from the captured original image of the electrode surface,and the ROI region which is extracted in the original image is the actual electrode region.Then the extracted ROI region is preprocessed by a proposed background standardization algorithm which is to suppress the gray value fluctuation of the pixels in the background region which is from the extracted ROI region image and eliminate the bright and dark stripes in the vertical direction of the ROI image.(2)In the defect detection part,an inspection algorithm combining rough defect detection and precise defect detection is exerted to inspect flaws in the preprocessed ROI region.The rough defect detection uses simple double-threshold segmentation algorithm and external rectangular transformation operation to quickly extract the potential defect region in ROI region image.Meanwhile,among the precise defect detection,an adaptive threshold segmentation algorithm based on gray histogram reconstruction is proposed to further judge whether the potential defect region extracted by rough defect detection is the actual defect region and accurately extract the corresponding actual defect region.The average IOU of the proposed adaptive threshold segmentation algorithm is 83.25% and the average running time of the algorithm is 0.022 s.(3)The defect classification part proposes a classification algorithm for electrode surface defects based on decision tree,which is used to classify the detected electrode surface defects.The average precise rate of the classification algorithm is 98.82%,the average recall rate of the algorithm is 98.80%,and the average running time of the algorithm is 139 ms.In addition,a parameter adjustment and optimization scheme of defect classifier based on feedback control is proposed.