中国科学院自动化研究所机构知识库

Point matching is a key step in robot writinglearning process. There are three major challenges that weakened most of the existing point matching algorithms to match points for Chinese characters, including nonlinearly deformation, connected strokes and geometrically dispersive. We propose a novel algorithm based on constrained global energy function (CGE) in the matching process to cope with the abovementioned challenges in this paper. We utilized a global spatial distribution energy function (EF) to evaluate relationship among point sets. Then we could solve problems with point registration by minimizing the energy function. To evaluate the matching results of energy function, we defined a vector that describe local spatial information to optimize the algorithm presentation. In addition, to avoid divergence, we designed an operator based on sigmoid function, using literation numbers as inputs to constrain the vector. We have conducted experiments on an extensive human handwriting database, and our algorithm performed competitively against the state-of-the-art point matching algorithm in terms of accuracy and stability.