基于自适应极坐标变换的景象匹配算法  

Exploring an Effective Scene Matching Algorithm Using Adaptive Polar Transform(APT)

作　　者：符艳军[1,2] 程咏梅[1] 潘泉[1] 李耀军[1] 孙开锋[3] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072 [2]空军工程大学电讯工程学院,陕西西安710077 [3]西安精密机械研究所,陕西西安710075

出　　处：《西北工业大学学报》2011年第5期702-708,共7页Journal of Northwestern Polytechnical University

基　　金：航空科学基金(20100853010)资助

摘　　要：对数极坐标投影变换以其对旋转及尺度的不变性而在景象匹配中广泛应用,但这种方法计算复杂且对图像的遮挡非常敏感。自适应极坐标投影变换,一方面保持了对数极坐标投影变换的旋转及尺度不变性,同时对图像间的局部遮挡具有鲁棒性且计算量相对较少。为此,把自适应极坐标投影变换应用到景象匹配中,提出了一种基于自适应极坐标变换的景象匹配算法。结合景象匹配辅助导航系统的实际给出了进一步的改进措施:根据自适应极坐标投影变换的特点,设计了一种新的相似性度量方法;创建虚拟圆索引模板,实现了在基准图上圆形子图的快速截取;在多分辨率匹配技术的基础上,进一步采用跳跃式搜索,有效缩减了匹配耗时。仿真实验表明,所提算法在实时性及鲁棒性方面优于传统的对数极坐标匹配算法。Aim. The introduction of the full paper reviews a number of papers in the open literature and then proposes the research mentioned in the title. Starting from Ref. 11, sections 1,2 and 3 explain our exploration. Section 1 briefs the APT of Ref. 11. Section 2 not only applies APT to scene matching but also deems it necessary to add three technical measures to make scene matching more effective than previously possible ： （ 1 ） a new similarity measure is designed using a threshold; （2） a fast circular subimage location method is used by creating an index template; （3） a pixel-jump searching method combined with a multi-resolution imaging technique is used to reduce the computational load. Section 3 analyzes the complexity of our scene matching algorithm and we believe we have made some progress in making it possess real-time capability. Simulation results, presented in Table 1, and their analysis show preliminarily that our scene matching algorithm is better than the conditional LPT based scene matching algorithm for real time and robustness.

关 键 词：对数极坐标变换 自适应极坐标变换 索引模板 象素跳跃式搜索 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]