【摘要】 针对传统动力电池焊点形态检测中存在的漏检、效率低和误检率高等问题，提出一种基于机器视觉的电池模组焊点缺陷实时检测系统.采用伽马变换、中值滤波处理，以增强图像清晰度与特征保留能力；提出一种改进的Canny边缘检测算法，以提高焊点边缘提取的准确性和完整性；改进Hu-SIFT快速匹配算法，以提升匹配的准确性；通过基于决策树的分类算法对焊点缺陷进行分类识别.试验结果表明，该方法能够精确识别焊点的形态特征，可对漏焊、虚焊、炸焊、焊点偏移及焊点氧化缺陷进行准确分类，平均检测准确率达到98%,符合电池模组生产线2组/s的节拍要求.更多还原

【Abstract】 In response to the challenges posed by missed detections, low efficiency, and elevated false positive rates associated with traditional inspections of battery cell weld point morphology, a real time defect detection system for battery module weld points based on machine vision has been proposed. The system utilized gamma transformation and median filtering methodologies to enhance image clarity while preserving essential features. An improved Canny edge detection algorithm has been introduced to further augment the accuracy and comprehensiveness of weld point edge extraction. Furthermore, a refined Hu-SIFT rapid matching algorithm has been developed to increase matching precision. A classification recognition method grounded in decision tree algorithms has been employed for the categorization of defects at the weld points. Experimental results indicated that the approach could accurately identify the morphological characteristics of weld points, thereby facilitating precise classification of defects such as insufficient welding, cold solder joints, excessive thermal damage, misalignment of solder points, and oxidation related complications. The detection system achieved an impressive accuracy rate of 98%, thereby fulfilling the production line requirement for processing two groups per second for battery modules.更多还原