计算机工程与应用 ›› 2013, Vol. 49 ›› Issue (14): 164-169.

• 图形图像处理 • 上一篇    下一篇

基于SURF特征的航空序列图像位置估计算法

乔奎贤1,赵  妮1,李耀军2   

  1. 1.西安工业大学 计算机科学与工程学院,西安 710032
    2.西北工业大学 自动化学院,西安 710072
  • 出版日期:2013-07-15 发布日期:2013-07-31

SURF-based position estimation method using aerial image sequences

QIAO Kuixian1, ZHAO Ni1, LI Yaojun2   

  1. 1.School of Computer Science & Engineering, Xi’an Technological University, Xi’an 710032, China
    2.School of Automation, Northwestern Polytechnical University, Xi’an 710072, China
  • Online:2013-07-15 Published:2013-07-31

摘要: 视觉传感器在航空无人机导航和定位任务中应用越来越广泛。针对无人机位置参数估计问题,提出了一种基于SURF特征的图像配准算法,该算法能够适应航空序列图像的旋转、尺度变换及噪声干扰,实现无人机位置的精确估计。构建了SURF尺度空间,运用快速Hessian矩阵定位极值点,计算出航空图像的64维SURF特征描述子;基于Hessian矩阵迹完成特征点匹配;使用RANSAC算法剔除出格点,实现位置参数的精确估计。通过航空图像序列实测数据位置估计实验,验证了该算法的有效性。

关键词: 视觉导航, 位置估计, 图像匹配, 快速鲁棒特征, 随机采样一致性

Abstract: Vision sensor is widely used in the aviation aircraft navigation and positioning tasks. For position parameter estimation of aircraft, this paper presents a SURF feature based image registration algorithm, it is able to adapt to aerial image sequence rotation, scale transformation and noise, to achieve accurate estimates of aircraft position. It builds SURF scale space, uses fast Hessian matrix maximum values to calculate 64-dimensional SURF feature descriptors of aerial images. Based on the trace of the Hessian matrix, it completes feature points matching task. By using RANSAC algorithm, it removes the outliers matching points for accurate position estimate. Experiments with two real aerial image sequences show the effectiveness of the proposed position estimation algorithm.

Key words: vision navigation, position estimation, image matching, Speeded-Up Robust Features(SURF), Random Sample Consensus(RANSAC)