CAJ | 학술논문

提出一种基于谱图理论的形状匹配算法。首先在表示形状的离散点集上构造待匹配形状的 Laplace矩阵，根据此矩阵的特征向量确定初始匹配关系，然后利用 Viterbi 算法将形状自身的结构特性(邻接性和弯曲能)融入求解过程，获得形状间的匹配关系。比较实验表明该文算法具有更高的匹配精度和鲁棒性。
제출일충기우보도이론적형상필배산법。수선재표시형상적리산점집상구조대필배형상적 Laplace구진，근거차구진적특정향량학정초시필배관계，연후이용 Viterbi 산법장형상자신적결구특성(린접성화만곡능)융입구해과정，획득형상간적필배관계。비교실험표명해문산법구유경고적필배정도화로봉성。
The author proposed a shape matching algorithm based on spectral graph theory. Firstly, we constructed the Laplacian matrices for the matched shapes represented by feature point sets, and the correspondences were estimated from the eigenvectors of the obtained Laplaican matrices. Then the structural characteristics ( adjacency and bending energy ) extracted from the shape were integrated into the solving procedure via the technique of the Viterbi algorithm so as to achieve the matching between shapes. Experimental results revealed the fact that the proposed algorithm possessed higher accuracy and better robustness.