基于广义立体像对的高分辨率遥感影像阴影检测

Shadow detection method for high-resolution remote sensing images supported by generalized stereoscopic image pair

由于高分辨率遥感影像存在同物异谱或异物同谱现象,常用的基于影像特征的阴影检测方法存在普适性较差、鲁棒性较低、高漏检率和误检率等问题,而基于物理模型的阴影检测方法受数据源限制应用较少。立足高分辨率遥感影像以及基础地理信息数据日益丰富这一背景,从摄影测量理论和阴影形成几何原理出发,提出了广义立体像对支持的、多源数据辅助的高分辨率遥感影像多角度阴影检测方法。实验结果表明,在地面控制点支持下,广义立体像对能够完成三维重建,且重建模型的定位精度能够达到影像分辨率对应成图比例尺的规范要求;该方法是可行的,虽然过程更加复杂,但由于理论上更加严密,能够避开常用检测方法存在的问题,取得更好的检测效果,具有良好的工程化应用前景。

The common shadow detection methods based on image features had certain problems such as poor universality,low robustness,high missing detection rate and high false detection rate,which caused by the phenomenon of same object with different spectrum or different object with same spectrum,while the shadow detection methods based on physical model are limited by data sources.Under the background of increasingly rich high-resolution remote sensing images and basic geographic information,a multi-angle shadow detection method supported by multi-source data and generalized stereoscopic image pair was proposed based on satellite photogrammetry theory and geometric principle of shadow formation.The test results show that:supported by a certain number of control points,the generalized stereo image pair can complete three-dimensional reconstruction,and the positioning accuracy of the model can meet the standard requirements of the image ground resolution corresponding to the mapping scale.This method is feasible,although the process is more complex,it can avoid the problems of common detection methods,and achieve better detection results because it is more rigorous in theory.It will have good engineering application prospect.