摘要
根据轮胎X光纹理缺陷区域灰度及灰度分布异常的特点,研究了一种通过分析统计特征进行在线缺陷检测的方法。在轮胎X光纹理灰度分布模型基础上,采用正则化预处理去除背景噪声,然后进行图像分块,分别计算每块的灰度均值和方差,并采用双线性插值运算形成均值图像和方差图像,再通过二值化实现缺陷检测。实验表明,与人工检测方法进行对比,该方法误判率低,检测精度高,并且运算速度快,能满足在线检测要求。
According to the abnormality of gray mean and gray distribution of defect region in tire X-ray texture images, an approach based on analysis of statistical teatures is proposed to implement on-line detection ot texture defects. Based on the gray distribution model of tire X-ray texture, the background noise of image is eliminated by normalization preprccessing. Image is divided into blocks, and then gray mean and variance of each block are computed. The mean image and variance images are formed by bilinear interpolation operation. At last, defect detection is achieved by binary processing of mean image and variance image. Compared to traditional manual detecting, experimental results show that lower rate of ntis-detection, higher precision and speed are achieved by this method to meet the demand of on-line detection.
出处
《光学技术》
CAS
CSCD
北大核心
2009年第1期60-62,66,共4页
Optical Technique
基金
国家科技支撑计划资助项目(2007BAF14B03)
关键词
纹理缺陷
统计特征
正则化
二值化
在线检测
texture defects
statistical features
normalization
binary processing
on-line detection
