Design And Implementation Of Real-time Monitoring System Of Book Bounding

As the time of Industry 4.0 is coming,industrial production has shifted from automation to intelligence.It is relatively mature that the robot arm performs the specified action to complete the industrial production and intellectualization is making great strides forward.Many fields at home and abroad have realized automated assembly line production operations,and automated operations have indeed greatly improved production efficiency and liberated manpower.In the process of book production and processing,from printing,to folding,to matching,to binding,all of the operations are done at once.How to detect the quality of book binding in real time under the requirements of high speed and high quality,and feedback the binding results,is of great practical significance for printing enterprises to improve production efficiency and reduce production costsFolding completed book post or leaflet by page number by hand or machine with integrated book is namely with the post.When matching,in order to facilitate matching and check the errors and omissions of matching,a small black box,namely ladder mark,is printed on the ridge of each printed sheet according to the order of the sheet.In the production process,there may be artificial or unavoidable matching failure,such as wrong,more,less and so on.Once it is found that the ladder logo image is out of order,it indicates that there is a matching error.Therefore,the detection of labeling image in the production process is related to the product qualification rate and production efficiency.In this thesis,aiming at the problem of quality detection of book post in the process of book production,a two-stage book matching detection algorithm is proposed,which uses YOLO to identify the ladder image and then judges by the regression algorithm,and a set of real-time monitoring system is implemented.The book falling will trigger the photoelectric switch fixed on the production line,and the photoelectric switch will trigger the industrial camera CCD to take pictures and take the image on the spine side of the book.CCD will capture the image through RJ45 to the computer.The computer sends a signal to the recycling device after the detection.The whole detection process is as follows: firstly,the returned image is clipped,the background area is removed,and only book part is retained.Then,the yolov3 model is used to recognize the clipped image,and all the ladder position information is recognized.Then,the midpoint of each ladder is extracted,and all the midpoints are regressed by the least square method.By judging the number of ladder marks identified,the fitting correlation coefficient,parameters and other information,we finally get whether there is a problem with book post.Through the test of local sample data and on-line test in the production environment,the detection time of each sample is within 700 ms,and the overall accuracy is more than 97%,which meets the demand of on-line production.