一种基于SVM的刹车蹄块片摩擦块表面裂纹检测法

A surface crack detection method for friction block of brake pad based on SVM

针对刹车片外观裂纹检测需求,通过构建刹车蹄块片图像采集系统,提出了一种基于支持向量机SVM的刹车蹄块片摩擦块表面裂纹检测法.该方法首先利用灰度图像的梯度模值信息,投影提取摩擦块所在区域;然后以局部窗口子图像为单位,计算灰度共生矩阵并提取相关特征量;最后采用分类样本对支持向量机分类模型进行训练,对摩擦块表面裂纹缺陷和正常区域进行分类预测.实验表明:该方法能较好地实现摩擦块表面裂纹缺陷和正常区域的分类,对表面裂纹缺陷存在与否的判定准确率可达98.33%.

Aiming at the need of surface crack detection of brake pad.an image acquisition system of brake pad is constructed,and a surface crack detection method of brake friction pad is proposed based on support vector machine(SVM).Firstly,the gradient modulus information of the gray image is used to project and extract the friction block region.Then,the gray level co-occurrence matrix is calculated and the relevant feature quantities are extracted based on the local window sub-images.Finally,the support vector machine classification model is trained with the classified samples to classify and predict the surface crack defects and normal regions of the friction block.Experiments show that the method can better classify the surface crack defects and normal areas of friction blocks,and the accuracy of judging whether surface crack defect exists or not can reach 98.33%.