论文部分内容阅读
针对传统局部不变特征的景象匹配算法冗余点多、实时性差、抗几何变换不突出的情况,提出基于CenSurE-star的无人机(UAV)景象匹配算法。首先采用Cen Sur E特征星型滤波器(CenSurE-star)提取基准图和实时图中的特征点,并生成FREAK二进制描述符;然后将汉明距离作为特征点的相似性判定度量,采用K近邻距离比值的方法提取匹配点对;最后利用基于RANSAC的定位模型得到空间几何变换关系,实现图像匹配并获取定位点经纬坐标。算法性能评价实验表明,本文算法不仅相对于SIFT、SURF、ORB算法,对各种变换具有更好的鲁棒性,而且相对于改进的SIFT、SURF算法处理时间有更大程度的缩短,算法定位误差在0.8个像素内,尺度误差在0.02倍内,旋转角度误差在0.04°内。基于算法进行外场飞行实验,实验证明算法定位精度较高,可以适应地貌信息较少的环境,并能满足无人机视觉辅助导航的需求。
In view of the fact that the traditional local matching feature matching algorithm has many redundant points, poor real-time performance and no anti-geometric transformation, this paper proposes a CenSurE-star UAV scene matching algorithm. Firstly, CenSurE-star is used to extract the feature points from the datum and real-time graphs, and the FREAK binary descriptor is generated. Secondly, the Hamming distance is taken as the measure of similarity of the feature points, and K nearest neighbor Distance ratio method to extract the matching point pairs; Finally, the use of RANSAC-based positioning model to get the geometric transformation of the relationship between the space to achieve image matching and latitude and longitude coordinates of the locating point. The algorithm performance evaluation experiments show that the proposed algorithm not only has better robustness than the SIFT, SURF and ORB algorithms, but also reduces the processing time of the SURF algorithm to a greater extent than the improved SIFT. The algorithm locates The error is within 0.8 pixels, the scale error is within 0.02 times, and the rotation angle error is within 0.04 °. Experiments show that the algorithm has high positioning accuracy and can adapt to the environment with less geomorphological information and can meet the demand of UAV visual assistant navigation.