无人机载框幅式高光谱影像的波段配准研究

[目的]高光谱成像技术在卫星遥感平台上由于飞行高度与技术限制等原因无法满足当前对于智慧农业与精准农业的需求,无人机平台的出现可以有效地弥补这一问题,给精细化、定量化研究农田信息提供数据支持。受平台稳定性,荷载能力等因素的制约,传统的推扫式高光谱成像仪不适用于无人机平台,而框幅式成像仪的应用前景较为广泛。但此类成像仪由于成像原理等因素的影响,获取的波段图像间存在姿态与位置差异,无法直接投入到后期应用中,在投入使用前需要进行波段配准。[方法]结合传统基于特征点的配准方法针对波段配准展开研究,对图像间灰度与位置差异对配准精度的影响进行了深入分析,针对传统的研究方法在匹配灰度存在非线性变换的图像上不足等问题,设计了两组匹配策略实验,分别为拍摄顺序的配准实验对比与波段顺序的配准实验对比。[结果]通过实验对比证明在引入拍摄顺序的变换基准的匹配策略下可以自动且稳定完成配准任务,此方法下配准精度可以达到亚像元级。[结论]文章所提方法兼顾了波段顺序与拍摄顺序,匹配结果可达亚像元级,是较优的匹配策略。

REGISTRATION OF FRAME HYPERSPECTRAL IMAGES BASED ON UAV

Hyperspectral imaging is a new technology for comprehensive and subtle cropland data collection and management. The UAV platform effectively surpass the traditional satellite platform in providing data for refined and quantitative research of cropland while the latter is strictly limited by orbital altitude and technological problems. On the UAV platform, frame imager is much better than traditional push broom imager, which is highly affected by platform stability and load capacity. However, registration is required before further application for band-spectral images of frame imager, which suffers from spectral and spatial discrepancy. Currently study on registration of hyper spectral data is not widely developed, due to its great spectral and contrast discrepancy. This paper developed an automatic and stable sub pixel registration method by comparing and analyzing multi band-images. And then it analyzed the impact of image gray scale and the position difference on the registration precision, designed two sets of experiment, i.e., order registration experiment and band sequence of registration. The results showed that the matching strategy can be achieved automatically and steadily under the matching strategy of the changing datum of the shooting sequence, and the registration precision can reach sub-pixel level by this method. The proposed method can take into account both the band order and the shooting order