The Ferrography Image Segmentation Based On Improved Grey Clustering

Wear often cause mechanical equipment failure and scrapped, lead to loss of energy andmaterials, resulting in a huge waste, endangering personal safety, and even the occurrence of a majoraccident. So talking wear condition monitoring and fault diagnosis of mechanical equipment hasimproved significant economic and social benefits. Ferrography is one of the effective technologywhich are extensively applied on wear particle analysis and machine condition monitoring. In recentyears, with the widely use of computers technology, image processing recognition and expert systemstechnology, Ferrography evolving towards intelligent direction. Ferrograph image segmentation is thefirst and crucial step of ferrograph technology. Achieving the accurate segmentation of wear debrishas an important impact on the extraction, recognition and statistical of abrasive wear characteristics.This paper briefly introduced the principle, characteristics and development status ofFerrography. Introduce the basic concepts of some commonly used methods of ferrography imagesegmentation, such as thresholding method, morphological segmentation algorithm and clusteringalgorithm. Gray theory in image processing applications is briefly introduced too.Aims at the accurate segmentation of wear debris chains and the abnormal large wear debriswhich are difficult to segment in ferrograph image, a new segmentation method by improved greyclustering is presented in this paper. First, subtract the background of ferrograph image with theimproved Otsu threshold method. Secondly, segment the ferrograph image which has subtracted thebackground with watershed algorithm. Third, the improved grey clustering with color characteristicsand relative position information of each region is applied to merge the over-segmented regions afterwatershed segmentation. Finally, base on the difference of shape between wear debris chains and theabnormal large wear debris to compare the equivalent ellipse axial ratio for identification andcorrection clustering results. The proposed algorithm is compared with Canny edge detection,K-means clustering and Fuzzy C-means clustering, the experiments show that the proposed methodis an effective way for the segmentation of wear debris chain and abnormal large wear debris.In this paper, Visual C++6.0is used as a software platform and OpenCV library used to fulfilledthe algorithms. The experimental results demonstrate that the methods of image segmentation andedge detection presented in this paper are feasible and effective.