Application Of Image Fusion In Detection Of Micro Bearing Surface Defect

Bearing is widely applied in machinery industry, and bearing quality concerns safety production problems and the usability and lifetime of machine. Bearing manufacturers care for quality test of bearing. Automatic production of bearing has been put into effect by means of technical innovation in recent years, but artificially assay is adopted in defect inspection and rejection of bearing. It is not only ill-timed with automatic production but also the reliability of the detection result is bad. Automatic defection detector is badly in need to solve the problem. It is testified that the indentation, scratch and other defections are random distribute on the outer-ring surface, seal ring and inside and outside of bearing.The inspection method of the defection would be improved and the reliability of bearing defect detection would be enhanced with this project carrying out and exploring. The CCD image recognition technique would be widely applicated in non-contact detection.For the micro-bearing, the image preprocessing and correction of image geometric distortion were the critical step of defect detection. Common image filtering methods were introduced in detail in the paper. The filtering method of image de-noising which was appropriate for the project was confirmed. In the image segmentation aspect, the threshold segmentation was introduced in detail. Hough transform Principle was used to find the center position of micro bearing images and separated detection targets from the original image. Basised on the method of geometric correction of image, the distorted bearing image was rectified.the definition and key technique of image registration were introduced in this paper, and the maximization of mutual information was adopted to rectify the bearing image. The image fusion technique was imported to the defect detection of the bearing for the first time. Key technique of wavelet analysis such as wavelet basis, decomposition, rules of low and high spectrum were made elaborate test.The fused bearing image was quantitative analyzed by the entropy, average gradient, spatial frequency and other evaluation index of the image.In this paper, the fused image was detected by the detection system which was designed by virtual instrument technique, the feasibility of testing program was proved.