Study On The Font Recognition Method Based On The Shape Similarity Distance

Handwritten Chinese character recognition is a very challenging project in the field of pattern recognition. It plays a great role in sorting letters, recognizing bank checks, managing statistical report and automatically inputting handwritten manuscript etc. Compared to other characters, handwritten Chinese character has characteristics of huge quantity, complicated structure, lots of similar character and serious distortion etc. Moreover, it is very difficult to accurately and fast identify the handwritten Chinese character.This paper based on the distance as the basis for judging the Chinese character similar degree, attempting to study and solve the problem of handwritten Chinese character in precise and accurate identification.There are four parts in this paper.It starts with overview of the current research status and the background of the application of Chinese character recognition. The pre-processing work on the images of handwritten Chinese character based on the HCL2000 by applying the image segmentation, two values, smoothing, thinning and normalization method was completed as well.Secondly, it fulfilled the experiments of reorganization based on centroid and FCM clustering in 3 data sets by taking different distances as the similarity measurement basis. Then the similarity of the performance of different distances is deeply analyzed and compared in this paper.Furthermore, a rapid system for handwritten Chinese reorganization was designed. It applied the elastic mesh technique extracted the feature vectors of Chinese character from the Chinese character library. A multi-template database with weight of Chinese character morphological feature was designed as well based on the clustering results of feature vectors. It applied the shape similarity distance as a classifier and the weight factors of morphological database were considered to complete the handwritten Chinese character recognition experiment.Finally, based on the software of Matlab, the simulation experiments were completed of the handwritten Chinese character rapid recognition system. The feasible and effective of this system were verified.