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构建  

Construction of intelligent detection model of electrode sheet surface defects based on adaptive theory and CART algorithm

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)

机构地区：[1]柳州工学院,广西柳州545005 [2]柳州汽车检测有限公司,广西柳州545005

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)。

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.

关 键 词：自适应阈值分割 CART分类 遗传算法 电极片 缺陷检测 

分 类 号：TP392[自动化与计算机技术—计算机应用技术] TP206[自动化与计算机技术—计算机科学与技术]