摘要
荧光磁粉检测技术是一种常见的无损探伤技术,常用于成品、半成品以及原材料的检验,以确保工件质量的可靠性。小波变换作为分析信号频率分量的数学工具,已成功地应用于图像处理的各个领域。将图像小波技术与磁粉探伤技术相结合,能快速、有效地提取出工件的缺陷特征。目前,磁粉缺陷检测通常通过肉眼观察,不仅速度慢,而且效率低。为了在工业生产中实现自动化、提高检测效率、拓展小波变换的实际应用,提出了一套基于小波变换的自动识别方法,并将其用于检测缺陷。首先,利用小波分析方法提取缺陷特征。图像经过小波变换后,可获得其低频系数和高频系数。高频系数较好地保存了图像的边缘信息(缺陷信息)。然后,对高频系数分量进行垂直投影等基础图像处理操作,以获取缺陷特征。最后,使用BP神经网络进行分类识别。测试结果表明,该系统对工件上的裂纹缺陷和凹坑缺陷具有良好的检测效果。
Fluorescent magnetic particle detection technology is a common non - destructive inspection technology ; it has been often used in inspection of finished products, semi - finished products and raw materials, to ensure reliability of the workpiece quality. Wavelet transform,as the mathematical tool for analyzing signal frequency component,has been successfully applied in various fields of image processing. The combination of image wavelet technology and magnetic particle flaw detection technology can fast and efficiently extract the defect features of workpieces. At present,the result of magnetic particle detection usually can be observed by naked eyes, which is slow and inefficient. In order to realize automation in industrial production, improve detection efficiency, and expand the practical application of wavelet transform, a set of automatic identification method based on wavelet transform is proposed to detect defects. Firstly, the wavelet analysis method is utilized to extract defect features. After wavelet transforming, the high frequency coefficient and low frequency coefficient of an image can be obtained, and the image edge information( defect information) can be better preserved in high frequency coefficient. Then, on the basis of high frequency coefficient component, the basic image processing operations such as vertical projection are conducted to obtain defect features. Finally, classifying recognition is realized by using BP neural network. The test results show that the system features excellent effects on detecting the crack defects and pit defects of the workpieces.
出处
《自动化仪表》
CAS
2017年第12期89-93,共5页
Process Automation Instrumentation
基金
国家"十三五"核能开发科研基金资助项目(20161295)
四川省重点实验室开放基金资助项目(13zxtk0504)
关键词
小波变换
特征提取
无损探伤
磁粉检测
BP神经网络
垂直投影
连通域
图像分割
Wavelet transform
Feature extraction
Nondestructive inspection
Magnetic particle detection
BP neural network
Vertical projection
Connected domain
Image segmentation