Research On Detecting Method Of Security Area Covered By Anti-counterfeit Paper Based On Machine Vision

As an application scenario of computer vision system,the print inspection has been developed and evolved for several years.The machine vision system also goes from packaging inspections to inspections of highquality-printing products or security printing products.In order to better control the quality risk of anti-counterfeit products,an improving solution for detection methods of safe line detection areas on passport-related printed applications was proposed.Using this method of promotion,full coverage of anti-counterfeit print inspections can be achieved to ensure product's quality and control the risk of defection.Anti-counterfeiting paper of passport adopts security features of allburied security thread.The security thread presents an elongated strip that runs through the entire piece of paper.At the same time,the color of the pattern area covered by the security thread will be slightly gray,and the position of the security thread placed in the paper is not fixed,resulting in the system mistakenly determining that the area covered by the security thread is a pattern defect,resulting in a lot of detection errors.To this end,a solution to false alarms caused by security lines is proposed.Based on the existing machine vision inspection platform,camera gain adjustments are made to increase the overall image acquisition brightness.According to the images collected by the platform,it is verified that the NCC(Normalized Cross Correlation)algorithm used in the existing machine vision inspection system cannot achieve the correct matching of the security line area under the reflected light.At the same time,the analogy and analysis of SURF(Speeded Up Robust Features)and SIFT(Scale-Invariant Feature Transform)are also performed to the performance of detection to the safe line region and verify that each algorithm has a detection bottleneck.In this paper,based on SURF feature points,the application of the HOG(Histograms of Oriented Gradients)description with rotation-invariance and low-dimension is proposed.Two solutions are proposed in this paper to improve its adaptability to rotating images.To verify that the HOG-SURF reaches the requirement of resisting the change of the illumination,the HOG-SURF descriptor is compared with the traditional SURF descriptor.The matching accuracy between HOG-SURF and traditional SURF was compared through a small batch test.The experimental results of HOG-SURF achieved the expected image matching effect.According to the accurate result of the security thread area matching,the security line locating algorithm is designed to accurately locate the security thread area.Based on positioning the line area,a method of image restoration was proposed.The restored image can be used as a subsequent defect detection for printed graphics.After testing,there is no obvious distortion in the restored image,and there is no excessive weakening of the detection capability for the security line area and its surrounding areas.This thesis implements and validates the feasibility and detection advantages of the HOG-SURF descriptor.The application of this descriptor can avoid the problem that the defect cannot be correctly detected due to the occlusion of the security line area.This method can improve the detection coverage of the product with anti-counterfeiting feature of security thread.