| 摘要: |
| 针对起伏较小三维地形的匹配系统，提出了一种匹配性能的快速估计方法。给出了匹配性能指标的定义，采用八联通区域的邻域完成了特征向量的多分辨率构造，使用多分辨率高斯马尔可夫随机场建立特征与匹配性能之间的关系。在特征提取的过程中，将基准图划分为若干个方块，提取的特征兼顾了邻域关系和多分辨率特性。给出了训练学习和推广的公式，使用最大后验准则，逐步推导出了性能指标估计值的解析解。最后，以实测数据为基础，通过实验验证了所提方法在运算效率和定位误差方面均有较好的性能。 |
| 关键词: 三维地形匹配 性能估计 高斯马尔可夫随机场 多分辨率 |
| DOI： |
|
| 基金项目:国家自然科学基金资助项目（41501362）；广东省公益研究与能力建设专项项目(2014A020218015)；广东省教育厅创新强校工程青年创新人才项目(15ZK0117) |
|
| A fast performance evaluation method for 3D terrain matching |
| SUN Sheng,LIU Renfeng,DENG Shaoping |
| () |
| Abstract: |
| A fast matching performance evaluation method is proposed for three-dimensional terrain matching systems with small-amplitude undulation. The definition of performance indicators for terrain matching is put forward at first. Eight-connected field is then employed to construct the multi-resolution feature vectors. Multi-resolution Gaussian Markov random field is used to model the relationship between terrain features and the matching performance.During the procedure of feature extraction,the reference map is divided into several rectangle blocks and both the neighborhood relationship and multi-resolution characteristics are utilized. Learning and testing equations are introduced and therewith the analytic solutions of estimation values for performance indicators are obtained by exploring maximum a posteriori criterion. Finally,it is validated that the new proposed method has better performance in terms of computing complexity and localization bias based on the ground truth sample data. |
| Key words: 3D terrain matching performance evaluation Gaussian Markov random field multi-resolution |
《电讯技术》编辑部 