针对SIFT算法处理遥感图像时存在计算量大、时间代价高的问题, 从极值点检测和相似性度量两个方面对SIFT算法进行优化改进。改进算法首先利用距离检测点越近的像素点对其影响越大的特点, 在极值点检测时选取距离检测点更近、权重更高的14个相邻点来替代SIFT算法中的26个邻域点, 减少极值检测的计算量。其次, 在SIFT特征向量匹配的相似性度量方面利用更简单的曼哈顿距离与切比雪夫距离的线性组合来替代欧氏距离, 减少特征匹配的计算复杂度, 提高匹配效率。最后通过实测遥感数据验证所提方法的有效性。

In processing high-resolution remote sensing images, the SIFT algorithm has large calculation burden and great amount of time cost.To solve the problems, we made improvement to the SIFT algorithm on extreme point detecting and similarity measurement.Firstly, the improved algorithm uses the 14 points that are closer to the detecting point and have higher weight instead of the 26 points of the SIFT algorithm to detect the extreme points, which can reduce calculation amount for extreme point detection.Secondly, in the similarity measurement of SIFT feature vector matching, the linear combination of Manhattan distance with Chebyshev distance is used instead of Euclidean distance, which is more simple and can decrease the computation complexity and improve the efficiency of matching.Finally, simulation results prove the effectiveness of the algorithm by using the measured remote sensing data.