国产高分系列遥感卫星影像的批量自动配准研究

Batch automatic registration of domestic Gaofen satellite images

关键词:  国产高分系列卫星  批量自动配准  AROP  中高分辨率影像

Key words:  domestic gaofen series satellites  batch automatic registration  arop  medium to high resolution images

摘要:

[目的] 遥感影像的精确几何配准对于国产高分系列卫星影像处理以及后续的农情监测等应用至关重要。[方法] 文章基于自动配准和正射校正程序包AROP，选择16 m分辨率的高分1号、高分6号宽幅影像和3 m分辨率高分2号、高分7号多光谱影像作为实验数据，以经过几何精校正的10 m分辨率的Sentinel-2影像和3 m分辨率的谷歌影像为基准影像，在农业用地场景和城市用地场景下，进行批量的几何配准实验，对配准的结果进行目视检验和定量评价。同时，为了更好地匹配高分影像与Sentinel-2影像，将高分1号、高分6号宽幅数据重采样到15 m的分辨率进行几何配准。[结果] 在农业用地和城市用地场景下，配准精度大多达到了0.5个像元以内，满足了影像配准的精度标准。其中15 m分辨率的高分1号和高分6号的配准精度分别为0.31～0.54和0.33～0.53，均小于0.6个像元；3 m分辨率的高分2号的配准精度在0.47～0.6之间，小于0.6个像元；高分7号的配准精度在0.44～0.49之间，小于0.5个像元，而且所有高分系列遥感卫星在配准后均目视效果出色，接边良好。[结论] 利用AROP对15 m分辨率的高分1号、高分6号宽幅影像和3 m分辨率的高分2号、高分7号的多光谱影像进行批量的自动配准，配准精度均在0.6个像元以下，符合应用的要求，减轻了劳动强度，提高了处理应用效率，为高分系列卫星的协同应用提供了技术支撑。

Abstract:

[Purpose] The precise geometric registration of remote sensing images is of utmost importance for the processing of domestically produced high-resolution satellite imagery and subsequent applications like agricultural monitoring.[Method] The article was based on the Automatic Registration and Orthorectification Package（AROP），selecting high-resolution images from GF-1 and GF-6 with a Wide Field of View（WFV）of 16 m resolution and multi-spectral images from GF-2 and GF-7 with a resolution of 3 m as experimental data. The Sentinel-2 images with a resolution of 10 m and Google Earth images with a resolution of 3 m were used as reference images. The study used geometrically corrected. The research conducted batch geometric registration experiments in agricultural and urban land scenarios. Additionally，in order to better match the GF images with the Sentinel-2 images，the GF-1 and GF-6 WFV data were resampled to a resolution of 15 m for geometric registration.[Result] In both agricultural and urban land scenarios，the registration accuracy mostly achieved within 0.5 pixels，meeting the precision standards for image registration. Specifically，the registration accuracy for resampled GF-1 and GF-6 of 15 m resolution was between 0.31-0.54 and 0.33-0.53 pixels，respectively，both less than 0.6 pixels. For GF-2 of 3 m resolution，the accuracy ranged from 0.47-0.6 pixels，under 0.6 pixels. GF-7 achieved an accuracy between 0.44-0.49 pixels，less than 0.5 pixels. Furthermore，all Gaofen series remote sensing satellites exhibited excellent visual effects and seamless edges after registration.[Conclusion] The use of AROP for batch automatic registration of resampled GF-1 and GF-6 WFV images at 15 m resolution，as well as multispectral images from GF-2 and GF-7 at 3 m resolution. The registration accuracy is all below 0.6 pixels，meeting application requirements. This research significantly reduces labor intensity，improves processing efficiency，and provides technical support for the collaborative application of the Gaofen Satellite series.