Texture Extraction And Identification Of Wear Particles By Combining Lifting Wavelet And Hough Transform

With the development of the technology of modern equipment, how to detect and diagnose the running status of mechanical equipment has become the concern of people nowadays. Ferrography is a kind of technology widely used in monitoring the status of the mechanical equipment and detecting the breakdown, in recent years, with the development of the computer technology, image processing technology and the intelligent technology, ferrography is becoming more intelligent. Ferrography image analysis technology is a fusion of ferrography and image processing technology and it is composed of ferrography image segmentation and wear particles analysis and identification which is the crucial step of ferrograph technology. How to get the feature of wear particles such as texture feature extraction is the premise of the accurate identification of wearing particle analysis, and the accurate identification of the type of wearing particle is the main basis of the status monitoring and fault diagnosis.The paper firstly introduced the principle, development and research status of ferrography, then introduced the mechanical wear mechanism and grain type, and some commonly used methods of image processing are introduced, such as median filtering, morphological processing and Otsu threshold, then introduced watershed segmentation algorithm and the texture research methods.To solve the difficult problem of texture extraction in ferrography, a new method for the texture extraction based on D4 lifting wavelet is presented in this paper(LWTE). First, a ferrography image is decomposed into RGB single-channel image, preconditioning each channel image by smoothing and morphology; then D4 lifting wavelet is used to each single channel image, which can get vertical, level and diagonal detail; and then the max-modular is used to extract the texture of the three channels and the final texture image is obtained through OR operation. Finally, verify the effectiveness of the method of the texture extraction by experiments.After the texture extraction, the texture identification of the wear particle based on Hough transform is presented(HTTI). The method aimed at the abnormal large wear particles such as severe sliding and fatigue wear particles, through the linear detection of texture by the statistical probability of Hough transform, can get the linear feature of the texture, like the length and angle; then get the shape of the abrasive characteristics, like the long axis and minor axis; then by analyzing the data can detect the linearity and directivity which distinguish between severe sliding and fatigue wear particles.At last, verified the effectiveness of the method. Results showed that the method can effectively extract the texture of wear particles and recognize the wear particles, which is an effective method of texture extraction and identification of wear particles.