基于低空摄影测量影像的特征点提取算子比较研究

发布时间：2018-06-30 00:22

  本文选题：低空摄影测量 + 特征点提取　； 参考：《地理与地理信息科学》2017年03期

【摘要】：影像匹配是低空遥感数据处理的核心步骤,而特征提取是影像匹配的基础。该文从兴趣算子的角度分析了摄影测量中几种主流特征点提取算法:Moravec算子、Forstner算子、SUSAN算子、Harris和SIFT算子,以角点类型较多的普通几何图形、卫星遥感影像和动力三角翼拍摄的低空遥感影像为数据,通过实验得到各算法的速度、精度、局限性和适应性。针对低空摄影测量影像,以重复率为指标定量比较和分析了各算法在抗噪、抗对比对变化、抗光照变化和抗旋转变化等方面的性能。实验结果表明,针对灰度信息丰富的低空遥感影像,SIFT算子具备尺度不变性,抗噪性最好,Harris算子提取速度最快,Forstner算子精度最高。实验结论为低空摄影测量影像处理提供了一种可行性方法。
[Abstract]:Image matching is the core step of low altitude remote sensing data processing, and feature extraction is the basis of image matching. In this paper, from the angle of interest operator, several main feature points extraction algorithms in photogrammetry are analyzed, such as: Moravec operator Forstner operator, Susan operator, Harris operator and sift operator. The satellite remote sensing image and the low-altitude remote sensing image taken by the power delta wing are used as data, and the speed, accuracy, limitation and adaptability of each algorithm are obtained through experiments. Aiming at low-altitude photogrammetric images, the performance of the algorithms in anti-noise, anti-contrast, anti-illumination and anti-rotation are compared and analyzed quantitatively with repetition rate as the index. The experimental results show that the sift operator is scale-invariant for low-altitude remote sensing images with abundant gray information, and the best anti-noise is the Harris operator, which has the fastest extraction speed and the highest precision of Forstner operator. The experimental results provide a feasible method for image processing of low altitude photogrammetry.
【作者单位】： 长安大学测绘与空间信息研究所;
【基金】：国家青年科学基金项目(41504001) 中央高校基本科研业务费(310826175027)
【分类号】：TP751
，

本文编号：2084019