The Application Of Image Processing And Target Detection In Passport Sewing Thread Detection

Passport belongs to confidential item,and it involves a lot of anti-counterfeiting techniques.Fluorescent line,with anti-counterfeiting function,is employed to sew and bind passport.After sewing and binding,sewing threads on passport show tight and dense arrangement of traces.At present,there is no appropriate method for the sewing thread extraction from passport to test binding quality of passport.This paper,through experiment,analyzes and compares the advantages and objective application conditions of several processing methods,based on existing research outcomes on image processing.With Matlab software,it solves the problem of detail loss in the process of image process,based on techniques like local homomorphic filtering in space domain,gray scale stretching,Butterworth filtering and so on.Also,it extracts binary image of sewing thread with sharp edges and contour and plump shape.By recognition and counting on sewing thread segments,the sewing quality of passport is determined based on the counting results.However,after statistical analys is of the experiment result,it demonstrates that false recognition of traditional image processing methods leads to false counting,which results in mistaken judges on test result.Detection accuracy of sewing thread acquired from this method is 32%,the consistency between judged result and real s ituation is 43%.To improve the recognition ability and recognition accuracy on sewing thread,this paper proposes a method based on Single Shot Multibox Detector(SSD for short)of deep learning convolution neural network.Targeting at the problem that dense sewing thread segments cannot be accurately located and recognized,this method marks the sewing thread image and constructs a dataset,and completes dataset training and prediction using multi-level feature image based on convolution neural network.And it finally obtains the sewing thread detection results through network test.The result shows that project of SSD object detection on passport sewing thread possesses the ability of effectively and accurately locating and recognizing sewing threading segments,and improves the detection performance.It achieves the detection accuracy of 84%,in a test conducting on 50 randomly acquired sewing thread images.Experiment result demonstrates that SSD object detection on passport sewing thread effectively solves false recognition problems existing in traditional image processing methods,and it is more capable of detection on dense arranged sewing threads.Through this paper's research on detection of sewing thread used in passport binding,it is beneficial to control passport binding quality during process.And it has practical s ignificance on sewing thread locating and recognition in passport binding.