基于爬山法的快速图像匹配

针对高速系统的应用需求,分析一种基于爬山法的快速图像匹配算法.文章首先通过分析NNPROD算法的相关度量曲面,设计了多个爬山者按照优先级依次爬山的启发式搜索算法,该算法能够保证全局最优.然后分析了进一步提高算法快速性的三种改进方案:爬山步长的自适应调整、通过图像样本集来计算相关度、爬山与分层相结合的搜索策略.实验结果表明,爬山搜索算法性能优越.目前,该法已在实际系统中得到应用.

Fast Image Matching Based on Hill-climbing

For a high speed system, a fast image matching algorithm based on Hill-climbing search is presented.By analyzing the correlative surface of NNPROD, a heuristic searching algorithm is defined for multi-climbers to climb the hill according to their priorities to ensure the global optimization.The paper also proposes three schemes to improve the speed of image matching, they are:adaptive adjustment of the climber's step, calculation of the correlative coefficient according to samples of the images, and a strategy of combining the Hill-climbing and Layered-searching.The experimental results show that the algorithms presented enjoy an excellent performance.