基于KD-Tree搜索和SURF特征的图像匹配算法研究

针对图像匹配时进行特征检测和匹配的搜索时间长的问题,文章研究了基于KD-Tree搜索和SURF特征的图像匹配算法。该算法首先提取得到图像的SURF特征并生成特征描述向量,然后为这些特征描述向量建立KD-Tree索引,最后通过计算每个特征点的与其距离最近的若干个KD-Tree上的最近邻点,完成特征匹配工作。实验结果表明,与SIFT算法相比,SURF算法进行特征检测的速度要快2～3倍;与全局最近邻搜索相比,基于KD-Tree索引的近似最近邻搜索大大减少了计算量,较大地提高了SURF算法的匹配速度。

Image Matching Algorithm Research Based on KD-Tree Search and SURF Features

According to the problem of long search time in detecting and matching features of image matching,so in this paper,an image matching algorithm based on KD-Tree and SURF features is researched.Firstly extract SURF features of images and create feature vectors,then build KD-Trees for these feature vectors,finally complete the image matching work by calculating the nearest neighbor vector which is nearest to each feature vector.The experimental results express that,the speed of detecting features by SURF algorithm is 2~3 times faster than by SIFT algorithm;in comparison with global nearest search,approximate nearest neighbor searching based on KD-Tree has Less calculation,so in this way,the proposed algorithm improves matching speed.