A Rifle Sight Based On Machine Vision Research On Defect Detection

Machine vision is a vision system that simulates human by machine,which makes the machine see and recognize the surrounding environment like human.At present,machine vision technology has been widely used in defect detection industry.Due to the limitations of the production process of the sight lens,there are four defects in the production process of the sight lens,namely black silk,black chicken silk,white chicken silk and spots.These defects have bad effects on the performance of the sight.At present,the defect detection of the sight lens is carried out by observing whether there is any defect through the microscope.In order to improve the detection accuracy and efficiency of the sight defects,this paper proposes a kind of aiming lens defect detection system based on machine vision,so as to realize the detection of the four defects of the sight lens: black silk,black chicken silk,white chicken silk and spot,instead of manual quality inspection,and reduce the labor cost.The main work of this paper is as follows.(1)In the detection of black silk and black chicken silk,the local histogram equalization of the image is carried out after the image is partitioned to reduce the influence of uneven illumination on the periphery and inside of the aiming lens.In this paper,a method of chicken wire location based on Hough line detection and regular hexagon translation is proposed,which solves the problem that the chicken silk line segments found by Hough line detection are incomplete and not ideal,and all chicken silk lines can not be found directly according to the existing line segments.An adaptive binarization method with feedback mechanism is proposed,which can automatically adjust the adaptive C value according to the feedback and realize multiple adaptive binarization of multiple images until the detection requirements are met.(2)After preprocessing,histogram equalization is used to improve the image contrast.After the image is down sampled,adaptive binarization is carried out to find out the white chicken silk network.Finally,the contour of the outermost edge is cut to leave the outline in the specified area,which is white chicken silk.(3)Based on the characteristics of the internal defects of the collimator,after preprocessing,the Gaussian filter is used to remove the Gaussian noise in the image,and the adaptive binarization is used to find the target contour,and then the image is opened to remove the fine and micro points on the image.In view of the filament interference,taking advantage of the fact that the ratio of the filament to its minimum circumscribed rectangle is far less than that of the spot to its minimum circumscribed rectangle,a threshold value is set for the ratio of the outline to the minimum circumscribed rectangle to remove the filament interference.The system is tested on the off-line machine with 40 sets of sight.There was no missing inspection of the product.The accurate recognition rate of the product is more than 93%.The reason of false detection is that white chicken shreds and black chicken shreds are in the critical state of judgment on many images,leading to false detection.In the test process,the detection system runs smoothly,without any abnormality or stuck phenomenon,which meets the manufacturer’s production requirements.