基于实时局部特征描述的自然路标提取与匹配

摘要

Local features extraction and matching techniques required in outdoor scene road sign matching are studied. The method in combination with curvature based feature extraction and binary feature description is proposed. The algorithm takes the advantage of reasonable features distribution in curvature algorithm to overcome the problems of traditional features uneven distribution. The real⁃time problem of feature matching is resolved by binary description algorithm. The validity and real⁃time per⁃formance of different algorithms are compared by experiments. The experimental results show that the proposed algorithm can ex⁃tract even features effectively while guaranteeing real⁃time performance.%针对户外场景路标匹配中所需的局部特征抽取和匹配技术进行研究，提出了基于曲率的特征抽取和二进制特征描述相结合的方法。算法利用基于曲率算法的特征分布较合理的特性，克服了传统特征分布不均的问题，且通过二进制描述算法解决了特征匹配的实时性问题，最后通过实验对比了不同算法的有效性和实时性。结果表明，该算法能够在保证实时性的同时提取出有效的均匀特征点。