A Detection System Of Printing Character Based On FPGA

At present,the regulation of assigned code is a trend of commodity management.Through the electronic supervision platform,the government can supervise the production,logistics,sale and use of the goods.As a kind of commonly used commodity regulatory code,the printing character of the detection and recognition is very important.For character recognition,the common method is mainly designed based on the software platform,which is not conducive to the intelligent management of the sited equipment.Therefore,this paper designs a set of printing character recognition system based on FPGA and Pulse Coupled Neural Network(PCNN)network.The main research contents and results are as follows:(1)In order to solve the problem that the traditional methods can't meet the demand of binary image segmentation,the PCNN algorithm combined with the minimum cross entropy theory is used to segment the image.First of all,through the analysis of the basic model of the PCNN algorithm,the image is segmented to binary image for the different characteristics of different gray pixel ignition time.Then,using the theory of minimum cross entropy between each segmented image and the original image is to make sure that which image is the best binary image by using PCNN to segment image.The method realizes the automatic segmentation of the image,and the effect is good.(2)For the problem that the traditional projection segmentation method can only segment continuous characters and can not segment the dot matrix character,this paper proposes an improved projection segmentation algorithm.The algorithm is to segment the character string to a single for extraction and recognition of characters by controlling the threshold size of segmentation and the relationship between the empty of printing dot characters and the character spacing.The improved algorithm not only can segment the continuous character,but also can solve the problem that the traditional projection algorithm can not segment the dot character.(3)For the feature of the printing dot character,this paper proposes an improved grid feature algorithm and redefines the recognition rules of characters.First of all,this algorithm dividing a single character into nine grids for character recognition by extracting features of each grid and combining with the recognition rules.Through the analysing the feature of the dot matrix characters,this algorithm can not only solve the problem that the geometric feature algorithm can not extract features of dot character,but also the algorithm can directly extract features without morphological processing.In addition,the algorithm is easier to implement for hardware than the neural network algorithm and the accuracy of can reach to 94%.The PCNN algorithm,the most cross entropy algorithm,the character segmentation algorithm and the grid feature recognition algorithm are implemented by FPGA,and finally,the key modules are analyzed by Modelsim software.The final results show that t each module is correct,and the feasibility of the system is verified.