Research On Feature Extraction And Recognition Of Hmong Character Image In Xiangxi

The digital protection of the Hmong Character in Xiangxi is a promising research topic,which can promote the promotion and dissemination of the characteristic traditional customs and cultures of the Hmong people in Xiangxi.Developing and researching the image recognition system of Xiangxi Hmong Character has certain practical value,and striving to improve the accuracy and recognition rate of the image recognition system is the goal of this paper.In this paper,based on different neural networks,a series of studies were carried out by using the collected image fonts of Xiangxi Hmong Character as research materials and using MATLAB programming.(1)In this paper,grayscale,binarization and stroke thinning will be used to make the image character more prominent and lay a foundation for recognition work.(2)In the aspect of statistical pixels,the distribution of macroscopic pixels of text is counted,and the distribution characteristics of the pixels projected on X and Y axes of text images are taken as the basis for recognizing characters,so as to achieve the purpose of recognizing characters.The optimal classification of axes is obtained by studying the projection partition of X and Y axes.(3)In the aspect of morphological structure,by analyzing the structure of Hmong Character,using a series of structural characteristics such as endpoints,forks,etc.,this paper extracts the representative feature matrix as the basis for recognizing characters,and improves the recognition effect by studying the means of dividing the text image blocks.(4)Based on the improvement of feature extraction methods in two aspects such as(2)and(3),the two features are combined to form a more complete feature matrix to improve the recognition accuracy of the system.(5)The recognition of block Miaowen image using BP neural network,PNN neural network and LVQ neural network is compared,through the experimental data charts,the advantages and disadvantages of these three kinds of neural networks for Hmong Character recognition are intuitively demonstrated.The difference of three kinds of neural networks in the underlying algorithm is analyzed.The respective characteristics of neural network classification algorithm and the recognition occasions they are good at.Combined with the characteristics of Hmong Character image in Western Hunan,the most suitable algorithm classifier in the experimental range is summarized.(6)The whole process of the experiment is reviewed and summarized.The future work tasks and improvement directions are prospected.It is hoped that further research will be carried out on the digitized recognition of the Xiangxi Hmong Character in the future.