Research And Implementation Of The Localization And Recognition Of Complex Distortion PDF417 Bar Code

PDF417 bar code as a two-dimensional code, because of its large storage capacity, error correction ability and other characteristics, has been rapid promotion. The decoding of bar code mainly includes three processes, including positioning, correction and identification. In this paper, combined with the structure characteristics of PDF417 bar code, the three processes of bar code decoding are analyzed and discussed. The main work is divided into the following several parts:First locate the bar code in the image. In this paper, the type of bar code distortion is divided into linear and nonlinear distortion, If the edge of distortion bar code is straight line, then it is linear distortion, otherwise is nonlinear distortion. Combined with the structure characteristics of PDF417 bar code, this paper presents a new method for locating. First cluster the rectangular blocks and the edge lines inside bar code, then matching the starting character and terminator to finish the bar code location. The method can locate the linear and nonlinear distortion PDF417 bar code.Then correct the bar code in the image. Use the traditional method of perspective transformation and bilinear interpolation to correct linear distortion bar code. In this paper, the nonlinear distortion is divided into two types: horizontal U-shaped distortion and vertical U-shaped distortion. For horizontal U-shaped distortion bar code, first use traditional correction method for correction, then block in the horizontal direction and correct each block, finally combine into a complete bar code. For vertical U-shaped distortion bar code, first according to the data block to block, then correct each block in tradition method, finally combine into a complete bar code. Experiments show that the method can go very well.Next recognize the bar code. This paper mainly study row and column segmentation of bar code. To solve the problems of traditional algorithm, this paper presents a method which combine projection and extracting peak by dynamic adaptive threshold values with distance detection to extract row boundary point. But for column segmentation, due to the possibility of uneven illumination, after using Sobel vertical edge detection, first make the bar code being binary, then combine projection and extracting peak by dynamic adaptive threshold values with distance detection to extract column boundary point, completing row and column segmentation.Finally use the modular design and MFC application framework to achieve the PDF417 bar code identification system. Then transplant the whole bar code recognition module to Linux system, and compile a bar code recognition library. Through a lot of testing, the system running results show that the system has good reliability and robustness.