Design Of Laser Anti-counterfeiting Code Segmentation And Recognition With Complex Background

Recent years, the phenomena of fake is always occurred in the manufacturing, it affects the development of economic seriously. Therefore, the automatic identification of product information and authenticity becomes more and more important in the modern production and sales. Information security technology is an effective security solution. However, due to the location of laser anti-counterfeiting code is not fixed, the background of laser anti-counterfeiting code is complicated. Traditional automatic character recognition can only recognize high quality characters with pure background. So the research to find segmentation and recognition algorithm for laser anti-counterfeiting code with complex background is of great significance.In view of the laser anti-counterfeiting code image with complex background, A character extraction method based on Mean Shift is proposed in the paper. Then three kinds of character segmentation algorithms are designed to segment normal characters, cross adhesion characters and overlapping characters. Lastly, LBP feature is applied to identify the anti-counterfeiting code and we get a high recognition accuracy. The main research work is as follows:(1) Studying the knowledge of Mean Shift. An improved Mean Shift algorithm is used to extract anti-counterfeiting code. The experiments show that the extraction algorithm we proposed can extract anti code from complex background. We also studied the knowledge of Hough transform and applied it to correct the anti-counterfeiting code image.(2) Studying the arrangement characteristics of anti-counterfeiting code characters. We divided anti-counterfeiting code characters into three kinds, normal characters, adhesion characters and overlapping characters. Three algorithms are proposed to segment each of them. The results of experiments show that the methods we proposed has more segmentation accuracy and recognition accuracy.(3) Studying the knowledge of template matching algorithm and LBP. LBP feature is used to identify characters. We compared its' results with the classification results of gray scale and grid features, it proved that LBP features can effectively resist the influence of complex background and damaged stroke in classification.Recognition rate of 97.69% are achieved with our method for laser anti-counterfeiting code. The experimental results show that our segmentation and recognition algorithm performs effectively for low quality images with complex background, fracture and deformation. The method has been applied in the laser anti-counterfeiting code identification terminals and had a good result.