Design And Algorithm Research Of Circulator Defect Detection System Based On Machine Vision

The manual inspection of wire in circulator has disadvantages such as poor repeatability and slow speed.Manual inspection method cannot meet the need of large-volume products.The application of machine vision technology in the field of defect inspection can improve the repeatability,speed and automation level of inspection.Therefore,focusing on machine vision and workpiece surface defect detection,we conducted a systematic study and algorithm design.Above all,We the discuss the lighting source,lighting method and image acquisition module of the system,Collect images in two ways: forward lighting and lateral lighting of the circulator.The image acquisition module is composed of a bi-telecentric lens and a 5million industrial camera to obtain high-contrast,low-noise images.Secondly,we start from the gray level difference of the background in the circulator,and divides the circulator to be detected into three parts: upper,middle and lower parts.For the three parts,we compare the contrast and sharpness of the images taken with front lighting and side lighting.In the end,we choose front lighting for the upper and middle parts,side lighting for the lower part.We analyze the gray and morphological characteristics of metal wires under different backgrounds and the types of edges that may be misjudged.Then,using dynamic threshold segmentation method to extract the edges in the image that are darker or lighter than the background.On the basis of skeletonization,we make the edges disconnected.the metal wire belonging to the ridge-shaped edge,the extracted edges are divided into two categories.According to the difference of gray-level features such as gray-level mean and gray-level deviation,we set multiple thresholds for the wire for further testing.Finally,using the machine vision defect detection system and metal wire detection algorithm designed in this paper,891 images of the circulator are detected,and the performance of the system is evaluated from the two aspects of false detection rate and missed detection rate.The experimental results show that this paper has good detection performance for the upper and lower parts of the circulator.Due to excessive noise,the missed detection rate and false detection rate of middle part are too high.In the end,the false detection rate of a total of 491 wires was 6.3%,and the missed detection rate was 6.9%.