| As an important means of storing and transmitting information, barcode has been widely applied in every aspect of our society. In recent years, the surge of internet-of-things has brought new opportunities for the development of barcode technology. Traditional one-dimensional barcode with its shortage of low information capacity, is no longer capable of meeting the actual demands, thus many two-dimensional barcodes have been devised. Among them, PDF417, with its unique advantages of high information capacity, wide aspect of encoding content and high reliability, has received world-wide acknowledgements and been frequently used. However, most of the PDF417 barcode readers already exist in the market, either with its technology been monopolized by abroad corporations, or is just capable of recognizing barcode in ideal image conditions. For the recognition of PDF417 barcode in various complicated image conditions, such as high noises, low contrast, distortion etc., there are still many problems to solve. Based on the research of its features, coding rules and error-correcting principles, this paper has done some work on the recognition method of PDF417 barcode under complicated background. The main innovations of this paper are listed as follows.1. To solve the problem of barcode localization in various complicated image conditions, such as high noises, low contrast, uneven illumination etc., a barcode localization method based on mathematical morphology and edge detection(MMED) is proposed in this paper. It firstly adopts some pre-processing techniques to the input image, then applies Canny edge detection and mathematical morphology operations to connect the discrete barcode region, and traces rectangular contours as candidate barcode area. Secondly, a line scanning strategy is employed to search for the unique pattern of start and end bar of PDF417, then some regional precision work is done to accurately localize the four vertexes of PDF417 barcode. Experimental results showed that the proposed method can effectively detect PDF417 barcode under various complex background.2. To solve the problem of distorted barcode localization, an improved barcode detection method based on MMED is developed, which firstly picks up convex quadrilateral contours and then applies different pattern searching strategy to achieve accurate location of PDF417 barcode. In addition, a concept of barcode rectangular degree is defined to distinguish two different distortion respectively caused by planar and three dimensional rotation. For barcode with the former skew, an affine transform is carried out. For barcode with the latter distortion, a perspective transform and bi-linear interpolation is utilized to achieve rectification. Experimental results show that the method can effectively detect and rectify barcode with different kind of distortions in real time.3. To overcome the disadvantages of traditional Sobel edge detection based column and row segmenting method, a new segmentation algorithm based on projection and waveform analysis is proposed in this paper. As for column segmentation, the binarized barcode image is projected along vertical direction and peak values are detected, then noises are eliminated through equal distance test. As for row segmentation, differences between each pair of adjacent rows are firstly calculated and peak values detection is implemented after some filtering process. Experimental results validate the effectiveness of the proposed methods4. An integrated system with relatively high recognition rate of PDF417 barcode recognition is designed and implemented. The system, with modularize structural design and OpenCV based implementation, has favorable expansiveness and real time performance. |
PDF Full Text Request