摘要
针对东巴文特有的结构特征进行了识别难点方面的分析,讨论了东巴象形文字的识别原理,提出了图像压缩、图像灰度化、平滑去噪、图像分割、归一化等图像预处理过程和基于特征提取的神经网络识别过程的东巴文识别方法,并通过实验对该方法进行了验证,实验结果表明了该方法的可行性。
The recognition obstacles of Dongba hieroglyphic are analyzed and the principle of recognition is discussed according to its special structure characteristics. Some methods used in the different processing phases are put forward. The image preprocessing including image compression, gray processing, smooth denoising, image segmentation and normalization is put forward. The method of feature extraction based neural network recognition is proposed. The experimental results prove the feasibility.
出处
《北京信息科技大学学报（自然科学版）》
2014年第3期72-76,共5页
Journal of Beijing Information Science and Technology University
基金
国家社会科学基金重大项目(12&ZD234)
关键词
手写体文字识别
图像处理
特征提取
神经网络
handwritten character recognition
image processing
feature extraction
neuralnetwork
