Research On Analyse And Recognition Of Particle Image On Digital Image Processing

Due to relative motion of two tribological surfaces, in the course of operation mechanical equipment produce plenty of wear particles which have diverse features because of the difference of wear mechanism. Ferrography has wear particle as its research objects, and through analysis of type and quantity of wear particle in the machine ferrography can do condition monitoring and fault diagnose of machine.With the development of computer science, it is an important direction to integrate ferrography with image processing, pattern recognition and artificial intelligence. To make ferrography keep up with the need of modem detection of mechanical state, we must research and develop intelligent system of condition monitoring and fault diagnose based on auto recognition of image for wear particle.Based on the review of achievement in the field of ferrography, especially in the application of image processing to ferrography analysis, this thesis have finished a lot of work including image processing, feature distilling and particle recognition. Firstly, we research novel Mean Shift filtering algorithm and RCA clustering algorithm through which wear particle image can be segmentation automatically, afterward we use mathematic morphology methods to separate wear particles from image background; Secondly, besides common methods for measure of particle features, we discuss fractal theory and apply it to measure features such as boundary details and surface texture of particle; Finally, based on improved BP neural network , the theory of gray relational analysis and optimize model between of them ,we realize automatic recognition of wear particle.