基于图像处理的轴承套圈裂纹检测系统

o solve the problem that the automation and efficiency of bearing magnetic detection is low, a crack detection system based on image processing is introduced. It includes median filter, band elimination filter, edge detection, PCA and SVM. The Deriche operator, which is used in edge detection, has higher accuracy and faster speed compared with Canny operator. Aimed at the problem that characteristic parameters and computation are large, the PCA is adopted to reduce the characteristic parameters from 8 to 3 dimensions. Then the difference of the computation time and accuracy, with or without PCA, is analyzed. Finally, the SVM is used to learn and classify the crack parameters. And results show that the accuracy is 97.3% when the kernel parameter is 0.02.