摘要
针对传统电极片表面缺陷检测精度低、检测效率不高的问题,提出构建一个基于自适应理论和CART算法的电极片表面缺陷智能检测模型。首先,采用基于灰度直方图重建的自适应阈值分割算法GHR-AT对电极片表面图像缺陷进行分割处理;然后通过遗传算法(Genetic Algorithm, GA)对分类回归树算法(Classification And Regression Tree, CART)进行优化,即利用GA算法代替二分法找到最优分裂点,以避免陷入局部最优问题,提升缺陷检测精度;最后将分割后的电极片表面缺陷输入至GA-CART算法中进行缺陷检测。实验结果表明,本模型对电极片表面图像缺陷检测精确率和召回率分别取值为97.84%和96.33%,相较于Faster R-CNN模型、SSD模型和Yolov5模型明显更高,且本模型的缺陷检测时长仅为7.62 ms,比另外三种模型分别低了10.05 ms、17.49 ms和23.84 ms。综合分析可知,本模型能够实现电极片表面缺陷的快速准确检测,检测效率显著提升,具备一定有效性和时效性。
Aiming at the problems of low defect detection accuracy and low detection efficiency,an intelligent defect detection model based on adaptive theory and CART algorithm is proposed.first,The adaptive threshold algorithm GHR-AT segmentation algo-rithm based on grayscale histogram reconstruction is used to segment the electrode slice surface image defects;Then by the genetic al-gorithm(Genetic Algorithm,GA)to classification regression tree algorithm(Classification And Regression Tree,CART)for the op-timization,That is,using the GA algorithm instead of the dichotomy method,To avoid falling into the local optimal problem,Improve the defect detection accuracy;Finally,the segmented electrode sheet surface defects were input into the GA-CART algorithm for de-fect detection.The experimental results show that the precision and recall rate of the electrode model are 97.84%and 96.33%re-spectively,which are significantly higher than the Faster R-CNN model,SSD model and Yolov5 model,and the defect detection time of this model is only 7.62 ms,which is 10.05 ms,17.49 ms and 23.84 ms lower than the other three models,respectively.Compre-hensive analysis shows that this model can realize the rapid and accurate detection of electrode surface defects,significantly improve the detection efficiency,and have certain effectiveness and timeliness.
作者
汤淑芳
龙鹰
阮威
彭梅
TANG Shufang;LONG Ying;RUAN Wei;PENG Mei(Liuzhou Institute of Technology,Liuzhou,Guangxi 545005,China;Liuzhou Automobile Tesing Co.,Ltd.,Liuzhou,Guangxi 545005,China)
出处
《自动化与仪器仪表》
2025年第2期18-23,共6页
Automation & Instrumentation
基金
广西高校中青年教师科研基础能力提升项目(2021KY1715)。
关键词
自适应阈值分割
CART分类
遗传算法
电极片
缺陷检测
adaptive threshold segmentation
CART classification
genetic algorithm
electrode chip
defect detection
