Correcting remote-sensed shaded image with urban surface radiative transfer model

Abstract

Shaded areas reduce the quality of digital images, bring many difficulties to image interpretation, especially in the high-resolution remote-sensed images. In order to improve the urban application capability, shadow correction should be performed. In this article, a new methodology for shadow correction was proposed applying the Urban Solar Radiative Transfer (USRT)11USRT: Urban Solar Radiative Transfer. model, which was a shadow correction scheme based on physical model. The sky view factor (SVF)22SVF: Sky view factor. was selected as one factor to quantify the morphological and structural characteristics of urban underlying surface, and then each radiation component was expressed quantitatively at pixel-level scale. Finally, the shadow correction was performed by calculating the actual reflectance of the urban materials in shadow areas. To validate the proposed scheme, three different spatial-resolution remote-sensed images were applied to perform shadow correction, including samples of Landsat-8, Sentinel-2, and Worldview-3 satellite image data. The results showed the following: (1) The shadow correction method can better improve the reflectance of the shadow area to the level of the illuminated area. There was a positive effect on the spectral reconstruction of different materials in the shadow areas after correction, and the spectral curves of corrected results coincided with the illuminated areas. (2) With the increase of spatial resolution, there was an increasing trend in the accuracy of shadow correction. In fact, the difference in the reflectance between the shadow corrected region and the illumination region decreased with the spatial resolution increasing. For example, after USRT shading correction, the rRMSE values of vegetation in the Landsat-8, Sentinel-2, and Worldview-3 images was 20.4%, 13.2%, and 9.8%, respectively. Generally, this proposed scheme can restore the spectral characteristics in the shadow areas and improve image quality, and there is a good application prospect in the field of high-resolution remote-sensed image processing.