Off-line handwritten Chinese character recognition method carrying out data expansion based on deformation method

Abstract

The invention provides an off-line handwritten Chinese character recognition method carrying out data expansion based on a deformation method. The recognition method comprises a step of building a platform on an image processor, wherein the platform is based on a Caffe deep learning framework including a plurality of types of convolution neural network models. Through elastic deformation, shearing and small-angle range rotation, a training data set is expanded, a test data set with tags is prepared, and the training data set is utilized to train the convolution neural network models on the image processor. The training data set is obtained by processing HCL2000 level-1 handwritten Chinese characters. Raw images of handwritten Chinese characters in an HCL2000 database are classified and subjected to elastic deformation, shearing and small-angle range rotation and then input into the convolution neural network models for network training. Finally, unknown Chinese characters are input into the models for testing and recognition results of the Chinese character images are obtained. The recognition method is highly intelligent, accurate in classified recognition, and fast in testing speed. The recognition method is good in recognition performance of low-weight databases and excellent in recognition capability of handwritten Chinese characters.

Description

technical field [0001] The invention belongs to the technical field of image classification, and relates to a method for classifying Chinese characters through data set expansion and based on a convolutional neural network model. Background technique [0002] The problem of offline handwritten Chinese character recognition has always been one of the difficulties in the field of pattern recognition research. Although a lot of related research has been carried out in recent decades, there are still many difficulties to be overcome in the problem of offline handwritten Chinese character recognition. These difficulties mainly come from the following aspects: large-scale Chinese character categories (only the first-level font library of GB2312-80 has 3755 kinds of Chinese characters), intricate Chinese character structure, a large number of similar Chinese characters, and the rotation or rotation of Chinese characters due to different writing habits. deformation. Some non-limit...

AI Technical Summary

Problems solved by technology

The above method has high requirements on the order of magnitude of the data set samples, and requires a large amount of data to train the network model, which does not work well for small orders of magnitude samples.

Moreover, the method is complicated. For different data sets, the effect of the model is somewhat different, and it does not have good universality.

Therefore, for the classification of small samples of offline handwritten Chinese characters, there is currently no method with strong recognition ability

Images