Feature Recognition Device Design Based On Image In Latex Condom

With people's living standards' improvement and the change of fertility concept,the demand for condom is increasing.In the first half of 2019,the annual production of condoms is close to 6 billion.According to the national GB / T 7544-2019,each condom must be subject to mandatory testing.At present,the factory test of condoms is mainly in electric dry test,which needs manual work.Due to the long work process and single action,the risk of workers getting tenosynovitis and cervical spondylosis becomes higher.In order to solve the problem of manual feeding in condom electric dry inspection,this thesis designed a set of automatic feeding device.The main contents are as follows:Firstly,the thesis designed a set of image-based condom feature recognition device to grab the tip of a condom,which was convenient for automatic feeding condoms.An automatic feeding device was built,which integrated posture adjustment,feature recognition,grasping and feeding.According to the operating steps,the PLC program was designed.Secondly,through the observation of the shape of latex condom,it was found that its characteristic area had a certain curvature profile,so the extraction and description of the outline of the condom were the basic work of the thesis.Through the comparison experiment and analysis of the current mainstream edge extraction operators,it was found that the Canny operator could get better edge information,and the detailed information was relatively complete,so the Canny operator was selected for edge extraction of condoms.Because the latex condom is soft and easy to produce wrinkles,there is often some line segment interference on its surface after contour extraction.Therefore,a contour retrieval method was proposed to remove the discontinuous and disordered line segments on the surface of the condom,and to improve the efficiency of subsequent algorithm recognition.Thirdly,in view of the characteristics of easy deformation and multi-rotation of the latex condom randomly set,it was proposed to use arc detection to locate the suspected tip feature area and do the mask operation.In order to facilitate the subsequent determination of the feature area,the polygon approximation method was used to process the suspected feature area and find the extreme points with high curvature on the contour of the mask area.Through the operation of these extreme points,the position of the feature area was finally determined.The feature area determination methods proposed in the thesis mainly included extreme point statistical method,angle statistical method,arch height and radius ratio determination method.When all three conditions in the mask area were met,the mask area was determined to be a feature region.Finally,the machine vision algorithm and the operation interface of the feature recognition device were programed.Using the software and the automatic feeding device validated the work of the thesis.The verification results show that the recognition rate and speed of the feature recognition algorithm designed in the thesis reaches the expected goal.