摘要
图像特征点匹配算法是增强现实几何一致性技术中的核心算法,目前图像特征点匹配算法耗时较大,准确性较差。提出了一种基于距离约束的改进SURF(Speeded-up Robust Features)算法:在特征点检测阶段,动态构建高斯金字塔图层,提高特征点提取的实时性和准确性;特征点的优化处理,避免提取到的图像特征点出现聚集现象。在特征点匹配阶段,对提取到的特征点构建KD-tree树索引,提高特征匹配的实时性和准确性。实验表明,改进的SURF算法有效地解决了目前方法存在特征提取时间相对较长,特征点匹配误差较大的缺点。
Algorithm for feature point matching is the core algorithm of geometric consistency technology in augmented reality area, while algorithms for feature point matching are time-consuming and have poor matching accuracy. An improved SURF （Speeded-up Robust Features） algorithm based on distance constraint was proposed. In process of detecting feature points, Gaussian Pyramid Layers was built dynamically to improve real-time and accuracy. And feature points were optimized to avoid feature points appearing on aggregation. In the stage of feature points＇ matching, the KD-tree indexes were constructed to improve real-time and accuracy. The experiment results indicate that the improved SURF algorithm effectively solves the problem of the current algorithm.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第12期2944-2949,2956,共7页
Journal of System Simulation
基金
"核高基"重大专项(2009ZX01038-002-002-2)
技部"原创动漫软件开发技术人才"计划扶持项目(2009-593)
