Research On Recognition Of The Wear Particle Based On Digital Image Processing Techniques

At present, wear particle recognition is mainly completed by manual work, the accuracy of which is too subjective too make further applications. The development of computer image processing techniques and artificial intelligence especially neural network created conditions for realizing intelligence of particle recognition, which have important significances for the study of auto wear particle recognition. This article is established in Ferrography and makes Ferrography analysis by computer image processing and artificial neural network techniques, which made wear particle recognition intellectualized and automatic. In this paper, the sorting of wear, the mechanics of wear particle formed and the characteristics of wear particles are introduced, and morphology characteristics parameter of wear particle are determined, which is particle brim digital feature. Based on image characters, after pre-process, section and extracting of contour parameter of wear particle using image process technique, four shape character parameters are extracted by fourier series expansion. After analyzing fundamental principle and shortcomings of neural network, current BP algorithm is improved in output optimization, network linearization implication optimization and adding momentum, and then astringe speed of BP algorithm. Auto wear particle recognition model is built. Wear particle recognition classifier is obtained according to four character parameters, and at last the classifier is trained and studied. The testing results of particle classifier using MATLAB tool cabinet, indicated that recognition ability of the classifier is 90%, and which can highly valued in application.