Research On Bearing Surface Defect Detection System Based On Machine Vision

Bearing is a critical component in reducing the friction of rotating parts.In the actual production process of bearing,various factors will lead to multiple surface defects,which will accelerate the damage of the bearing and lead to equipment failure.The existing bearing surface defects detection is mainly by manual visual inspection,so it is challenging to meet the national demand for high-quality bearing products.Therefore,it is urgent to develop the bearing surface detection technology with fast detection speed and a high degree of automation.The research focuses on bearing surface automatic detection technology and its system development based on machine vision technology.The main findings are as follows:A set of bearing surface defect detection systems based on machine vision was designed according to the practical application demand of bearing surface defect detection.The system determined the mechanical structure of the detection module,the robot vision equipment,and the fixed installation mode of the bearing.Through the whole system circuit analysis and the local hardware designed,the control circuit design of the detection system was completed,and the control software of the detection system was developed based on X-Sight software.Given the poor accuracy of bearing surface defect location,we proposed an edge feature extraction method based on image recognition.In the process of filtering,by comparing the edge detection experiments of the Sobel operator and the Canny operator,the edge detection of the Canny operator was optimized by using a non-maximum value to suppress interpolation and an iterative method to select a high threshold,and the optimal strategy to select low threshold,which improved the accuracy of defect location.To identify and segment the types of bearing surface defects,several kinds were detected and identified.The image was filtered,binarized,and morphology processed by extracting the ROI of the defect area so that the bearing surface defects can be located.For bearing black skin defects,an image segmentation method based on the Otsu algorithm with weighted class variance was proposed,which improved the accuracy of threshold and detection efficiency,and achieved better image segmentation results.For the bumping defects,the dynamic threshold method of local threshold segmentation was used to realize the defect segmentation,and the effect meets the experimental requirements.Finally,the defect region processing and different feature extraction were realized using morphology to segment the image.Finally,the detection system’s control software was developed using X-Sight software.The software studied the control method of robot motion trajectory and completes the recognition and sorting of bearing surface defects.The relative experiments were designed and compared with manual detection.The method improved the stability and reliability of the detection process.The proposed algorithm improved detection accuracy and sensitivity compared to the traditional Canny algorithm.