基于机器视觉的复杂背景下车轮踏面损伤检测

Wheel Tread Damage Detection in Complex Background Based on Machine Vision

复杂背景加大了车轮踏面损伤监测的难度,为了提升检测效果,降低检测误差,提出一种复杂背景下车轮踏面损伤机器视觉检测方法。估计车轮踏面图像的噪声方差,利用小波系数分别对图像不同区域展开去噪处理。根据灰度共生矩阵提取车轮踏面损伤区域的纹理特征;采用K-Means++聚类算法聚类处理踏面子区域的纹理特征,并通过损伤区域轮廓梯度信息检测损伤轮廓点,同时引入形态学方法,合并损伤区域,最终实现车轮踏面损伤机器视觉检测。结果表明,所提方法可以获取精准的车轮踏面损伤检测结果。

The complex background increases the difficulty of monitoring wheel tread damage.In order to improve the detection effect and reduce detection errors,a machine vision detection method for wheel tread damage under complex backgrounds is proposed.Firstly,the noise variance of wheel tread image was estimated,and then the wavelet coefficients were used to remove the noise from different regions of the image.Secondly,the texture features of the wheel tread damage area were extracted based on the gray level co-occurrence matrix.Thirdly,the K-Means++clustering algorithm was adopted to cluster the texture features of sub regions of tread.Meanwhile,the damage contour points were detected by the gradient information of the damage contour.Moreover,the morphological method was introduced to merge the damage regions.Finally,the machine vision detection of wheel tread damage was achieved.Simulation results show that the proposed method can obtain accurate detection results of wheel tread damage.