【摘要】 现有计算机在指导手写汉字练习与测试中，指导依据大都基于全局特征，缺少基于更细粒度特征。提出一种基于局部信息的手写汉字笔画提取方法，为手写汉字评价与指导等任务提供数据支持。首先提取出汉字骨架并对骨架中的毛刺与断裂等问题进行优化；然后使用PBOD算法提取汉字交叉区域后，对交叉区域进行合并删除，消除笔画形变，通过局部信息计算笔画段的组合系数，根据组合系数提取笔画；最后根据获取到的笔画数与算法迭代次数，动态调整组合系数阈值，保证在正确连接笔画段的前提下提取到更可能多的笔画。将该方法在手写汉字数据集上进行实验，其准确率、召回率与F1值分别达到了95.91%、95.71%与95.81%，可用于后续的手写汉字评判与指导等任务。更多还原

【Abstract】 In the practice and test of handwritten Chinese characters, most of the guidance bases are based on global features, but lack of more fine-grained features. This paper proposes a stroke extraction method for handwritten Chinese characters based on local information, which is used to extract the local features of handwritten Chinese characters′ styles and features, and provide solid data support for tasks such as handwritten Chinese character evaluation and guidance. First, extract the Chinese character skeleton and optimize the burr and fracture in the skeleton. Then, use the PBOD algorithm to extract the cross regions of Chinese characters, merge and delete the cross regions to eliminate stroke deformation. Connect the corresponding stroke segments according to the combination coefficient to extract strokes, Finally, according to the number of strokes obtained and the number of algorithm iterations, the combination coefficient threshold is dynamically adjusted to ensure that more strokes can be extracted under the premise of correctly connecting stroke segments. The method is tested on handwritten Chinese character dataset, and its accuracy, recall and F1 value are 95. 91%, 95. 71%and 95. 81% respectively. This method can be used for subsequent tasks such as handwritten Chinese character evaluation and guidance.更多还原