Research On Condom Automatic Feeding System

As common sexual health products,the demand for condoms is increasing day by day.According to the 2019 new condom national standard GB/T7544-2019,every condom must be tested for pinholes.At present,workers need to repeat the mold operation to cooperate with the dry method electrical inspection machine in the process of quality inspection.In response to this problem,this thesis studied and designed a set of automatic condom feeding device based on machine vision to solve manual nesting operation in the condom electric inspection process.The main contents are as follows.Firstly,according to the working principle of the condom dry method electrical inspection machine,the thesis gave a system scheme of the condom automatic feeding device.The system was mainly composed of a manipulator,an unrolling condom platform,a set of PLC,a Raspberry Pi,and an industrial camera.The manipulator completed the work of reclaiming and transporting the condom.The condom platform transported the condom to the image acquisition area.The industrial camera transmitted the collected condom pictures to the Raspberry Pi.The Raspberry Pi performed the feature recognition on the location of the condom seminal vesicles.Secondly,because of the complex working environment on site,the condom image collected by the industrial camera would have certain noises.The thesis used the median filter of the adaptive filter core.In addition,due to the special contour shape of the condom,it was necessary to perform edge detection on the condom image before feature recognition,and then execute contour extraction.In the aspect of edge detection,the thesis used the improved Canny operator to solve the problem that the traditional Canny operator would blur the image boundary and manually setting the double threshold.Because condoms were easily deformed,there would still be a large number of irregular line segments after edge detection.To solve this problem,the thesis applied an edge contour extraction method that could be to any pixel image to remove a large number of useless line segments.It built a foundation for subsequent feature recognition basis.Thirdly,the thesis performed the feature recognition of the seminal vesicle position on the condom image after contour extraction.Because the contour of the characteristic area had special arc characteristics,the thesis used the Hough gradient method to locate all possible characteristic areas in the initial positioning stage,and then masked and fitted polygon curve processing to study the related laws of the extreme point values,angles and contour shapes in the mask area and summarize the methods for judging the characteristic areas of the seminal vesicles in condom pictures with different pixels to realize condom feature recognition.Finally,according to the relevant configuration of the Raspberry Pi,the thesis used Python to write feature recognition software based on the Open CV vision library,and sent the condom feature location to the PLC through the communication protocol.Experiments show that the recognition rate of the seminal vesicle region of any pixel condom image used in this paper has a recognition rate of 96.2%,and the recognition time is 60?260ms when using a 0.3 or 0.5megapixels industrial camera,which basically achieves the expected goal.