Microcrack feature extraction method for chaotic optical surface of silicon nitride ceramic bearing roller based on multi-scale wavelet transform enhancement and optimized PSO-FCM coupling

To address the issue of chaotic optical gradient feature and non-target interference factors affecting the feature extraction effect of chaotic optical surface microcracks in silicon nitride ceramic bearing rollers, an algorithm for extracting microcrack features from chaotic optical surfaces of silicon nitride ceramic bearing rollers based on multi-scale wavelet transform enhancement and optimized PSO-FCM coupling is proposed. Through the db4 basis function of multi-scale decomposition of microcrack features, the soft threshold function is constructed to deeply denoise the image. The normalized fusion features of microcracks after multi-scale vector decomposition are enhanced, and the gradient information is enhanced while retaining the microcrack features on the chaotic optical surface. The particle swarm optimization mathematical equation with a decay function is built to optimize the FCM chaotic clustering model. The nonlinear decayed particle velocity equation is constructed to update the particle position and iteratively refine the optimal clustering center positions to realize the feature extraction of microcracks. The experiment showed that the reinforcement index RESE and PSNR in the chaotic optical surface microcracks of silicon nitride ceramic bearing rollers reach 15.76 and 24.19, respectively, effectively overcoming the problems of chaotic features in the surface microcracks while retaining the defect features. The segmentation indices Miou, F1 score, accuracy, and recall of the optimal clustering center [231 161] reached 0.912, 0.972 08, 0.998 32, and 0.985 60, respectively, overcoming the influence of non-target interference factors and chaotic optical gradient features to achieve complete feature extraction of surface microcracks.