一种基于改进地貌形状上下文的形状匹配方法

发布时间：2018-10-09 14:05

【摘要】：在基于地貌形状上下文的形状匹配方法中,计算地貌空间测地距离消耗时间较高,对应形状特征提取过程的效率较低.针对这一问题,本文提出了一种基于地貌模糊形状上下文的快速形状匹配方法.在形状特征提取过程中,通过引入最短路径算法对轮廓采样点间的测地距离进行快速计算.在此基础上结合对数极坐标模糊直方图构造地貌模糊形状上下文,其能够更好地描述轮廓点分布情况进而有效提升形状描述符的表达能力.考虑到轮廓点集顺序已知,进一步引入动态规划分析不同地貌空间下形状片段间的对应关系,以获取准确的形状匹配结果.通过对不同的数据集进行实验仿真分析,验证了本文方法能够有效地提升运算效率并取得较好形状检索精度.
[Abstract]:In the shape matching method based on the geomorphological shape context, the computation of the geomorphological spatial geodesic distance consumes more time, and the extraction process of the corresponding shape features is less efficient. To solve this problem, a fast shape matching method based on geomorphological fuzzy shape context is proposed. In the process of shape feature extraction, the geodesic distance between contour sampling points is calculated quickly by introducing the shortest path algorithm. On this basis, combining logarithmic polar fuzzy histogram to construct geomorphologic fuzzy shape context, it can better describe the distribution of contour points and effectively enhance the expression ability of shape descriptor. Considering that the sequence of contour points is known, dynamic programming is introduced to analyze the corresponding relations between shape segments in different geomorphological spaces in order to obtain accurate shape matching results. Through the experimental analysis of different data sets, it is proved that the proposed method can effectively improve the efficiency of operation and achieve better shape retrieval accuracy.
【作者单位】：大连理工大学电子信息与电气工程学部;
【基金】：国家自然科学基金(61374154) 中央高校基本科研业务费专项资金(DUT16RC(4)18)资助~~
【分类号】：TP391.41

【相似文献】