提出了一种基于棋盘距离的骨架定义方法，此骨架结构更接近了区域的实际形状，骨架的计算更为简单，传统的利用骨架信息识别目标的方法仅使用骨架的结构信息，而本文提出了骨架结构信息和骨架深度信息相结合的目标识别方法，在骨架的基础上提出了数种形态特征，以这些特征作为神经网络的输入进行目标识别得到较好的识别结果。

In the paper, A new skeleton definition of region based on chessboard distance is submitted, which is more close to the true figure of region than traditional skeleton, the algorithm is brief. Instead of only using the structure information of skeleton in traditional recognition method based on the skeleton information, The depth information and the structure information of skeleton are combined in the algorithm. A series of new feature to recognize the target is submitted in this paper also. The feasibility of the algorithm is proved by the experiment results.