行文工整的甲骨拓片文字特征提取算法研究

RESEARCH ON FEATURE EXTRACTION ALGORITHM FOR CHARACTERS ON ORACLE BONE RUBBINGS IN METICULOUS AND NEAT WRITTEN STYLE

针对行文较工整的甲骨拓片,通过对现有的粗网格特征提取算法与点特征提取算法的研究与改进,提出一种新的网格点特征提取算法。该算法将点特征引入到粗网格特征中,把绝对地址变成相对地址,将点特征网格化,并将位置关系融入特征向量。使这两种算法互为补充,相得益彰,大大提高了识别效果。网格点特征提取算法,对行文工整的现代汉字的手写体文书的识别效果也很好。

Based on research and improvement of existing coarse mesh feature extraction algorithm and the point feature extraction algorithm, a new grid point feature extraction algorithm is put forward in this paper for the oracle bone rubbings with the characters in meticulous and neat written style. In this algorithm the point feature is introduced into the coarse mesh feature, the absolute addresses are converted to relative addresses, and the point features have been made gridded and the position relationship is integrated into the feature vector. These two algorithms are complemented each other and get along well for more outstanding benefitS, thus the recognition effect is improved significantly. The grid point feature extraction algorithm also has good recognition result on the handwritten documents of modern Chinese in meticulous and neat written style.