Abstract: A new optimal model based on connected components is proposed，with which the breadth of parts，single characters and joined characters can be self-adaptive confirmed.In addition，weighted k-mean method is used to divide joined characters.The whole method not only improves self-adaptive ability，but also improves the veracity of segmentation.By analysis，this method has satisfactory performance for segmentation of handwritten Chinese character.

Key words: connected components, least squares, unconstrained, handwritten Chinese character segmentation

摘要： 在连通域单元的基础上提出了一种手写体汉字切分的优化模型，该模型可以自适应的确定部件、单个字和粘连字的宽度。另外，对粘连字的切分采用了加权k的均值法。整个切分方法既提高了算法的自适应能力，又提高了切分的正确率。实验表明这种方法具有很好的切分效果。

关键词: 连通域, 最小二乘, 无约束, 手写体汉字切分