Research On Target Location Method Based On Varistor Image

The quality of the varistor directly affects its performance.Computer identification and human visual recognition are not able to look at the quality of the varistor from the internal structure,so the defect detection on the appearance is very important.A good method of locating objects in an image is a necessary step for defect detection.In order to locate and segment the varistor image more accurately to realize the automatic construction of the varistor image data set necessary for deep learning.This paper presents a method for positioning and image segmentation of varistor body and pins based on Hough transform and mathematical morphology.According to this topic,the main work of this article is as follows:(1)We had used the laboratory-provided coaxial light source device to collect front-side images of varistor images,and had obtained images that eliminate surface reflections.Because the varistor has good quality and inferior quality,we had collected two types of varistor images in the end.(2)First,we had pre-processed the image of the varistor image acquired,and preprocessed the image based on denoising,graying and binarization.We had then used Hough transform based on circle detection to locate the resistive body.In order to further separate the main body and the pins,we first perform an edge search on the positioned main body part,and then fill the background of the main body.Finally,we had used mathematical morphology to remove the edge marks of the subject to locate the pins.(3)The final experimental positioning and segmentation show that the actual results of the proposed method are ideal and have a good target segmentation effect,which is beneficial to provide reliable varistor image data sets necessary for deep learning.