基于机器视觉的电极片缺陷检测方法研究  

Research on Electrode Flaw Detection Method Based on Machine Vision

作　　者：吴中虎 胡伟[1] WU Zhong-hu;HU Wei(School of Electrical Engineering and Automation,Henan Polytechnic University,Jiaozuo 454003,China)

机构地区：[1]河南理工大学电气工程与自动化学院,焦作454003

出　　处：《印刷与数字媒体技术研究》2025年第1期101-109,共9页Printing and Digital Media Technology Study

基　　金：国家自然科学基金(No.U1804147)。

摘　　要：本研究将机器视觉与锂电池模切设备相结合,提出一种基于Canny算法改进的极片缺陷边缘检测方法。首先采用灰度拉伸的自适应图像增强算法,增强图像信息;其次针对对比度较低的缺陷使用引导滤波对图像进行降噪,有效避免了高斯滤波在降噪时对梯度变化平缓的边缘造成梯度消失的问题;采用基于Sobel算法的3×3梯度模板使用一阶有限差分计算4个方向偏导数阵列得到梯度的幅值与方向角;最后,基于迭代法和NIBLACK算法自动获取图像的高、低阈值,通过双阈值去掉伪边缘,并将两种算法的检测结果逻辑与运算进行边缘融合。经实测,本算法对极片表面低对比度的缺陷边缘提取识别率可达97.95%,相较与传统Canny算法提高了17.35%以上。且该算法系统可以在线稳定运行,满足实际生产需求。A method for edge detection of electrode defects based on Canny algorithms improvement was proposed by combining machine vision with lithium battery die-cutting equipment in this study.Firstly,an adaptive image enhancement algorithm based on grayscale stretching was adopted to enhance image information.Secondly,directional filtering was used to denoise images with low contrast defects,effectively avoiding the problem of gradient vanishing caused by Gaussian filtering on edges with gentle gradient changes during denoising.Using Sobel algorithms based 3×3 gradient template used first-order finite difference to calculate the amplitude and direction angle of gradients for four directional partial derivative arrays.Finally,automatic acquisition of high and low thresholds for images was based on iterative method and NIBLACK algorithm.Pseudo edges were removed through dual thresholds,and the detection results of the two algorithms were logically and operationally fused for edge fusion.Through actual testing,the algorithm proposed in this study has achieved a recognition rate of 97.95%for edge extraction of low contrast defects on electrode surfaces,which is more than 17.35%higher than the traditional Canny algorithm.And the algorithm system can run stably online and meet actual production needs.

关 键 词：机器视觉 引导滤波 CANNY算法 缺陷检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]