Research On Magnetic Column Surface Defect Detection Method Based On Image Processing

In the process of magnetic column forming,sintering and grinding,surface defects such as scab,black flake,edge drop and crack will appear on the surface of magnetic column.The existing human eye detection is subjective and the defect classification standard is inconsistent,so the traditional manual detection method is no longer applicable.Aiming at the problems of low efficiency and easy to be affected by human,this paper uses the image processing method to detect the surface defects of magnetic column,designs the defect detection and recognition system,and focuses on the segmentation,recognition and classification of the surface defects of magnetic column.The main work is as follows(1)A defect detection and recognition system is designed.Through the selection of optical system and camera,the basic experimental equipment and conditions of image acquisition part are determined,and a new image acquisition method is described.(2)Hough transform is used to extract the defect region of magnetic column image.Histogram equalization method is used to enhance the contrast between the defect area and the background area of the magnetic column image.According to the noise distribution characteristics of the magnetic column image,the mean filtering and median filtering are analyzed experimentally,and the median filtering method is determined as the denoising method of the magnetic column surface image.(3)Two image segmentation methods,threshold segmentation and edge detection,are used.In the thresholding segmentation,2D-Otsu image thresholding segmentation method is used to improve the performance of 2D-Otsu algorithm by adding weight coefficient;and artificial fish swarm intelligence algorithm is introduced to improve its sensing range and moving step size.The effectiveness of the improved artificial fish swarm algorithm is verified by performance test experiments.On this basis,a 2D-Otsu image thresholding segmentation algorithm based on dlafsa is proposed,which improves the performance of the algorithm The artificial fish swarm algorithm combined with threshold segmentation can effectively detect image defects through comparative experimental analysis.In the edge detection,the mean parameter of the maximum inter class method is improved,the high and low thresholds are improved by adding the improved maximum inter class variance method,and the canny edge detection operator is improved.The experimental results show that the 2D-Otsu image threshold segmentation algorithm based on dlafsa has fast segmentation speed and good segmentation effect for obvious magnetic column surface defects,and the improved Canny edge detection operator can completely detect the complex edges of image defects,which improves the detection accuracy of the algorithm.(4)The binary image of the magnetic cylinder surface defect image is segmented.The geometric and shape features are used to extract the features.The area,width and aspect ratio feature parameters are used to distinguish the defect types,which can effectively distinguish the scar defect and the crack defect.The effect of the black chip defect and the edge drop defect is not ideal.The method of ellipse fitting is used to determine the distance between the black chip defect and the edge drop defect The center pixel coordinate position is used to distinguish the black chip defect from the edge drop defect.Finally,using the difference of characteristic parameters of different defects,the improved BP neural network algorithm is used to recognize and classify the magnetic column defects.The experimental results show that the recognition and classification accuracy of the improved BP neural network algorithm reaches 95.58%,and there is no missing detection.The method has high ability to identify the surface defects of magnetic column,and the detection speed and accuracy can meet the actual detection requirements.