| As one of the important high-tech information age, the character recognition has been applied to many fields, and become one of the most active in the field of pattern recognition. This article is mainly concerned about the auto-recognition of printed capital English letters and digits in background image. Character recognition technology consists of three modules in general, those are: image pre-proeessing module, characteristic extraction module and character recognition module. Algorithms of the modules related to character recognition are deeply studied and analyzed in this paper, According to the character of the complicated background of real characters, the technology of image processing, chaos and artificial neural network is used integratedly. And, a series of algorithms used in character recognition based on our research are finally settled down.In the vehicle character images preprocession, the technology of second global threshold segmentation is used, the paper presents a gray image binarization method by which characters and the background in the image is effective segmented, solve the interference problem which is existence of the characters goal background and get a good segmentation effect. Second threshold algorithm is extremely simple, two threshold values are obtained automatically by using the global threshold technique, avoiding waste of time of the slow local threshold, while overcoming a problem which is caused a single threshold segmentation ,it is that background is as target or target points are included in the background wrongly, lay the foundation for accurate character recognition. In the part of character segmentation, this paper has segmented the character image into a single character by using the improved horizontal projection, and normalized the character after segmenting.Character recognition stage, characteristic extraction is in progress by using improved rough grid feature, and directly input the unified character primitive feature to BP neural network classifier to recognize character. In the design process of Classifier, this paper uses the technology which combines chaos optimization algorithm and gradient descent algorithm of BP neural network, image characters classification and recognition method based on chaotic optimization BP neural network is presented to complete the network parameters optimal design, An effective solution to avoid falling into local minimum frequently and low convergence speed in network training, moreover, suitable network structure is selected through repeated trial and comparison, finally we recognize the charaeters smoothly.Based on research of the theory, this paper has developed an automatic recognition system modeling for character and realized the corresponding algorithms using Matlab programming tool. |
