A feature extraction method for silicon nitride bearing microcracks based on adaptive gamma and multi-width dynamic thresholds

This study aims to address the problems of localized indistinctness, low contrast, and unclear feature recognition in images of silicon nitride bearing surfaces by proposing adaptive gamma variation and multi-breadth dynamic threshold segmentation, achieving accurate identification and extraction of localized weak features of microcracks on silicon nitride bearing surfaces. To analyze the local weak features of microcracks on the bearing surface of silicon nitride, we construct the gamma correction function equation, define the appropriate pixel value to enhance the contrast of the image, design the dynamic threshold calculation function equation, define the local area with different breadths, and set the optimal threshold value to achieve the accurate segmentation of the local weak features. The experimental results show that the average peak signal-to-noise ratio of the enhanced image of local weak features of microcracks on the bearing surface of silicon nitride reaches 48.80 db, and the pixel accuracy of the feature extraction image reaches 90.0%, which effectively enhances the local weak features of microcracks in silicon nitride and optimizes the correctness of the local weak feature extraction of microcracks in silicon nitride.