Research Of Laser Anti-counterfeiting Code Recognition Under Complex Background

In recent years,fake products have become a major public nuisance in the development of market economy.Anti-counterfeiting technology is a measure to protect the brands,markets and consumers.Optical character recognition technology has been very mature now,but these are limited to high quality characters with pure background.However,character recognition rates are lower due to change of background or introduction of noise or influence of non-uniform illumination.So the research of laser anti-counterfeiting code recognition system under complex background is of great significance and commercial value.In view of the laser anticounterfeiting code image under complex background,the character pre-processing and recognition is researched in the paper.The main research work is as follows:(1)Studying the knowledge of Maximally Stable Extremal Regions(MSER),a text detection algorithm is proposed,which employs an improved MSER algorithm as basic letter candidates.These candidates are then filtered using geometric and stable region information to exclude non-text objects.One feature that separates anticounterfeiting code from other elements of an image is its nearly constant stroke width.This can be utilized to extract anti-counterfeiting code based on an improved stroke width transformation method.We compared its results with the extraction results of Mean Shift algorithm and K-means algorithm,it proved that our approach can effectively resist the influence of complex background and non-uniform illumination in character extraction.Moreover,to improve the quality of the image,a series of pre-processing steps had to be performed including filtering,binarization,morphological operation and tilt correction based on Radon transform.(2)Accurate location of character is recognized by horizontal projection and prior knowledge.we propose a method for character segmentation,firstly,the project profile histogram method is employed to obtain the no touching or overlapping characters from the separated blocks of the characters string.Then,for the touching characters,the segmentation is performed by introducing minimum weight segmentation path algorithm and using the concave and convex of string profile.Experimental results show that the proposed method are better than drop fall algorithm and vertical projection algorithm.(3)Studying the knowledge of Support Vector Machine(SVM)and Histogram of Oriented Gradient(HOG)feature.HOG feature is used to identify characters.We compare its' results with the classification results of Local Binary Pattern(LBP),grid features,Geometric features and fusion features.In addition,the proposed approach is compared with some existing methods such as template matching technique and neural network classification method.it proves that HOG features can effectively resist the influence of complex background and damaged stroke in classification.The proposed method can improve the accuracy of character recognition effectively,and the algorithm has a lower time complexity.Recognition rate of 98.75% are achieved with our method for laser anticounterfeiting code.The experimental results show that our pre-processing and recognition algorithm performs effectively for low quality images in complex background.The method has been applied in the laser anti-counterfeiting code identification system,which shows its effectiveness.