Research Of Video Image Characters Recognition Algorithm And Its Implementation On DSP

The characters of video image have an important role in understanding the content of the image and extracting characters from video images contributes to the analysis and processing of the image information. Video image characters recognition technology involves digital image processing, computer graphics, pattern recognition, computer vision, artificial neural network and many other theories and methods.This paper simulates median filtering algorithm with different sizes of the median filter templates based on MATLAB and selects3×3size of the median filter template. Through the analysis of the text frame detection algorithm of frame difference method, this paper puts forward the frame edge difference algorithm. This paper combines multiresolution decomposition, sobel edge detection, horizontal and vertical corrosion operation, dilation operation with multi-resolution fusion as the text localization algorithm. According to the text characteristic, this paper sets some text screening rules and the experimental results show that the text localization algorithm gets very great effectiveness.This paper simulates and analyzes different text segmentation and different tilt correction algorithms. According to the simulation results, this paper adopts k-means clustering to segment the text image and Hough transformation to correct image tilt. This paper segments the characters based on the projection of characters and this paper improves the method to apply to connected characters. In order to improve the recognition rate, this paper normalizes the coordinate recognition characters to standard size. This paper uses BP artificial neural network to recognize the characters. In order to get a better recognition network, this paper simulates basic BP algorithm, vary learning rate BP algorithm and adding momentum factor BP algorithm. By analyzing the training time and convergence, this paper combines vary learning rate BP algorithm with adding momentum factor BP algorithm. The experimental results show that the recognition rate of this algorithm can achieve94%above.This paper uses ICETEK-DM6437-B of Realtime Technology CO. as the developing hardware platform and implements the video image characters recognition algorithm on the platform. This paper uses C language to complete the video image characters recognition algorithm in the integrated environment of CCS. Firstly set the related environmental parameters of BIOS and creat related objects, then save the configuration file and load the file into the project and then write C codes, finally modify the C codes until meeting the requirements. In order to improve speed of the algorithm, this paper also optimizes the mathematical morphology operator, the characters segmenting algorithm, the definition of the variables and the loops.The experimental results show that the implementation of the video image characters recognition algorithm on DSP basically meets the expectation and this paper provides experimental basis for the futher research and development.