| 2D Barcodes have the advantages of high density, large information capacity, high reliability, excellent capability of error correction, ability of storing multiple information, high confidentiality, ability of anti-counterfeit and low cost. In recent years,2D Barcodes, like other automatic identification technologies (e.g., radio identification technology, RFID), have been applied widely in many fields including national defense, customs, taxation, public safety, transportation, post, medical treatment and so on.Though deep study into the recognition technologies of 2D Barcode, this thesis designed and developed a 2D Barcode recognition terminal based on DSP. By considering the critical technologies of the 2D Barcode (e.g. Data Matrix) recognition, in terms of image processing and barcode localization, some core algorithms were proposed in this thesis. Moreover, a complete automatic recognition algorithm for 2D barcode was derived by integrating the core algorithms as well as the necessary ones of geometry transform, decoding, and error correction. This complete recognition algorithm was used in the corresponding product design.With respect to image preprocessing, two innovative algorithms for image enhancement and binarization were proposed respectively:1. A fuzzy reasoning based rule for wavelet domain image fusion was drawn, and an image-enhanced algorithm was designed which fused barcode images based on this rule. This algorithm solved the problem that the contrasts of barcode image in parts were too low to decode the barcodes correctly in practical applications.2. A binarization scheme was proposed. This scheme eliminated the effect of uneven illumination with the illumination distribution estimated by wavelet analysis, and then binarized the processed image using the Otsu method. This scheme solved the decoding problem caused by the poor binarization effect, which was a general result of uneven illumination or too complex background in industry applications.With respect to barcode localization, this thesis proposed two novel algorithms for barcode region extraction and precise localization respectively:1. A Data Matrix region extraction method, namely GF-BPNN, based on Gabor filtering and BP neural network was proposed, solving the problem of completely and accurately extracting the barcode region from complex background. 2. A Data Matrix barcode localization method jointly based on edge tracking and Radon transform was proposed. This method localized the barcode precisely and overcame the existing disadvantage of the barcode-localized method which detected line edge using Hough or Radon transform (i.e., the difficulty of finding a balance between computational complexity and localization precision).In addition to the algorithms about the two aspects of 2D barcode recognition technologies aforementioned, this thesis further designed and developed a new high-speed 2D barcode recognition terminal based on DSP. The hardware platform of the terminal consisted of a data processing subsystem which adopted high performance floating point DSP TMS320C6713 as the core processor, and an image capture subsystem with FPGA as the controller. The corresponding software was based on the innovative image preprocessing and barcode localization algorithms above. These innovative algorithms and self-designed hardware platform constituted this DSP-based embedded 2D barcode recognition terminal with independent intellectual property rights. Field experiments in industry producing application demonstrated that this terminal could recognize different kinds of 1D/2D barcodes (e.g. 2D barcodes of Data Matrix and PDF417, 1D barcodes of Code39 and Code128) in the production lines fast and correctly. Both of its recognition speed and accuracy have reached domestically leading and internationally advanced level.
|
PDF Full Text Request