用于快速形状匹配的精确型高度函数特征描述

发布时间：2018-05-13 19:26

本文选题：形状匹配 + 精确型高度函数　；参考：《光学精密工程》2017年01期

【摘要】：在形状匹配过程中为了提升高度函数描述子的检索精度和对边界噪声与局部变形的鲁棒性,本文提出了一种精确型高度函数特征描述算法。首先提取目标形状外轮廓,构造轮廓采样点的精确型高度函数描述子并进行特征降维,接着利用优化后的并行动态规划进行形状匹配,最后引入形状复杂度分析提升匹配效果。基于点的几何特征显著性,提出形状精度理论,进一步分析局部形变与边缘噪声对形状特征描述的影响。在MPEG-7数据库、Swedish Leaf数据库、Tools数据库和ETH-80大型3D数据库上进行匹配实验以及在Kimia99数据库上进行抗噪实验,实验结果表明:本文提出的算法效率高,匹配时间仅为高度函数描述子的12.5%,在MPEG-7和ETH-80上的检索率最高分别为90.38%和90.07%;在Swedish Leaf和Tools上,检索精度最高分别为95.07%和94.86%,检索性能和鲁棒性均优于高度函数和其他重要算法;在添加噪声的Kimia 99上,该算法的抗噪性能优于高度函数描述子,即使在噪声水平为2.0的情况下,依旧能保持91.92%的检索率。本文提出的算法检索精度高,效率高,鲁棒性好,抗噪性强,具有较好的可扩展性,能有效地应用于形状检索领域。
[Abstract]:In order to improve the retrieval accuracy of height function descriptor and its robustness to boundary noise and local deformation during shape matching, an accurate height function feature description algorithm is proposed in this paper. Firstly, the external contour of the target shape is extracted, the precise height function descriptor of the contour sampling point is constructed and the feature dimension is reduced. Then the shape matching is carried out by using the optimized parallel dynamic programming. Finally, the shape complexity analysis is introduced to improve the matching effect. Based on the salience of geometric features of points, the theory of shape accuracy is proposed, and the influence of local deformation and edge noise on shape feature description is further analyzed. The matching experiments are carried out on the MPEG-7 database, the Swedish Leaf tools database and the ETH-80 3D database, as well as the anti-noise experiments on the Kimia99 database. The experimental results show that the algorithm proposed in this paper is highly efficient. The matching time is only 12.5 of the height function descriptor, and the highest retrieval rates on MPEG-7 and ETH-80 are 90.38% and 90.07%, respectively, while on Swedish Leaf and Tools, the retrieval accuracy is 95.07% and 94.86%, respectively, and the retrieval performance and robustness are superior to the height function and other important algorithms. On the noisy Kimia 99, the performance of the algorithm is superior to that of the height function descriptor, and the retrieval rate of 91.92% can be maintained even when the noise level is 2.0. The algorithm proposed in this paper has high retrieval accuracy, high efficiency, good robustness, strong anti-noise, good scalability, and can be effectively applied in the field of shape retrieval.
【作者单位】：湖北工业大学机械工程学院;
【基金】：国家自然科学基金资助项目(No.51675166,No.51205115)
【分类号】：TP391.41

【相似文献】