将目标形状的轮廓看成一个无序的点集，从中抽取形状特征，用于快速而有效的目标识别是形状分析任务中的挑战性问题.针对该问题，提出了一种基于复杂网络模型的形状描述和识别方法.该方法提出用一种自组织的网络动态演化模型构成一个分层的描述框架，在网络动态演化的每一个时刻，对网络分别进行局部测量和全局测量，抽取网络的无权特征和加权特征.在形状匹配阶段，用获得的局部描述子和全局描述子分别进行局部匹配（基于Hausdorff距离）和全局匹配（基于L₁距离），组合两种匹配的距离值构成对形状的差异度度量.用标准的测试集对所提出的方法进行性能测试，实验结果表明，所提出的算法能够快速而又鲁棒地完成较高精度的形状识别任务.

Treating the shape contour as an unordered point set and extracting shape features from it for fast and effective shape recognition is a challenge task of shape analysis. To address this issue, a complex-network based shape description and recognition method is proposed in this paper. In this method, a self-organized dynamic network-evolution model is built for providing a hierarchical description framework. In each moment of the dynamic evolution of the complex network, local and global measurements are performed against the network shut that both weighted and un-weighted features are extracted from the network. At the shape matching stage, the local matching (based on Hausdorff distance) and global matching (based on L₁ distance) are conducted using the obtained local descriptor and global descriptor respectively. The dissimilar value between two shapes is determined by combining the two distance measures. Several standard test sets are used to evaluate the performance of the proposed method, and the experimental results show that the proposed method can provide robust and fast shape recognition in high accuracy.

国家自然科学基金（61372158）；江苏省自然科学基金（BK20141487）；江苏省“333工程”高层次人才资助项目（BRA2015351）；江苏高校科研成果产业化推进工程项目（JHB2012-18）；江苏高校优势学科建设工程资助项目（PAPD）；江苏省政策引导类计划（产学研合作）-前瞻性联合研究项目（BY2016009-03）