中文摘要：

姿态角测量误差是导致机载遥感成像精度低的主要原因之一．为提高机载遥感成像的精度，并针对当前姿态角的测量误差较大、高精度惯性导航设备价格昂贵等问题，提出了一种应用基于SURF的航摄图像匹配的算法来解算飞行平台姿态角的方法．首先采用多线程对相邻航摄图像并行提取SURF特征点，然后使用改进的基于KD-tree的近似最近邻匹配方法寻找相邻航摄图像匹配的特征点对，并应用RANSAC算法剔除错误匹配点对，再通过最小二乘法得到图像的变换模型参数，最终解算出机载平台的姿态角变化量．对该方法的实现进行了详细的比较分析，仿真结果也证明了该方法的有效性．

英文摘要：

Attitude measurement error is one of the main factors that deteriorate the imaging accuracy of airborne remote sensing. In order to improve the imaging accuracy of airborne remote sensing, and in view of the fact that high-precision inertial navigation instrument are very costly, a low-cost, an improved image matching methods based on Speeded Up Robust Features （SURF） algorithm is proposed to calculate the attitudes of the aircraft platform. Firstly, SURF algorithm is used to extract series of the interest points of two adjacent aerial images. Then the improved nearest neighbor matching method based on KD-tree was used to get the pairs of the matching interest points. And the RANdom SAmple Consensus （RANSAC） al- gorithm is used to remove the error pairs of matching interest points. Finally, the transformation matrix between the two selected images is established to obtain the attitude angles. The realization of the method is specifically analyzed. The simulation results also demonstrate the effectiveness of the method.