| Lead-acid battery is a commonly used rechargeable battery widely applied in the fields of automobiles,UPS power supply,solar power generation,etc.Due to its crucial importance,the production process of lead-acid batteries requires strict quality control and inspection.Defects are inevitable in the production process,and if not detected and handled in a timely manner,these defects can cause a decrease in battery performance,shorten its lifespan,and even pose safety risks.Therefore,developing an efficient and accurate defect detection method is of great significance to improving the production quality and safety of lead-acid batteries.The aim of this research is to explore an online defect detection method for manufacturing electric vehicle batteries based on machine vision technology.This aims to solve the current difficulties in defect detection during the battery manufacturing process,and improve the production efficiency and quality of the battery production line.To achieve this goal,this paper utilizes various machine vision technologies,including image preprocessing algorithms,image segmentation algorithms,image filtering algorithms,and morphological processing algorithm modules,to detect potential defects in the manufacturing process,such as plate defects and welding defects.Additionally,a machine vision-based detection system has been developed to provide hardware and software support for implementing this method.The main work content is as follows:1.The aim of this article is to construct a visual system framework for the development of an overall detection system.The construction of the system includes the selection of industrial cameras,lenses,and light sources,as well as the capture of images.In terms of image processing,this article utilizes overexposure technology for image preprocessing according to the characteristics of the defects to be detected.In software development,Open CV and C++ are used for mixed programming to develop image processing software,enabling automatic defect detection.Additionally,for convenience of operation and data visualization,a human-machine interaction interface is developed using Qt,ultimately realizing the development of the overall detection system.2.This study proposes a new detection algorithm for plate defects.The main contributions include: an adaptive dual-region growing algorithm for background removal,segmentation-based enhancement of images to address unstable and non-smooth enhancement effects,a segmented weighted least squares method designed by incorporating battery characteristics for angle estimation to address issues such as high calculation complexity and low accuracy in estimating tilt angle,and finally,a dual-threshold method used on the traditional OTSU algorithm to improve the accuracy of plate extraction.3.This paper proposes a welding defect detection algorithm.Based on the Sobel operator,edge extraction is performed and non-maximum suppression is applied to the fitted edges to remove the background.To solve the problem of paper separation segmentation adhesion,the traditional mean filtering algorithm is weighted iterated.The segmented image is binarized using the Otsu algorithm,and boundary extraction and hole filling algorithms are used to compensate for the reflective areas.Finally,the morphology algorithm and Hough circle detection are used to determine the welding defects.After completing the above tasks,corresponding on-site experiments were conducted using this system.The test results show that the average accuracy of the polar plate defect algorithm is about96%,and the time for the polar plate image to be processed from acquisition to completion is about51 ms.The average accuracy of the welding defect algorithm is about 97%,and the processing time for a single welding image is about 61 ms,which meets the requirements of on-site practical testing. |
