ICARE: A physically-based model to correct atmospheric and geometric effects from high spatial and spectral remote sensing images over 3D urban areas

Automatic mapping of urban materials from remotely sensed radiance images remains difficult because of the complex phenomena induced by relief. Indeed, shadows and environment effects disrupt the radiance reaching the sensor. The measured radiance also depends on the illumination conditions of the observed area. This paper describes a new physical model, ICARE, able to solve the radiative transfer inversion problem in urban areas, in the reflective domain (0.4-2.5 mu m), from high spatial and spectral resolution images. This new approach takes into account the relief, the spatial heterogeneity of the scene and atmospheric effects, in order to extract rigorously the ground surface reflectance. The resolution method consists of modelling separately the irradiance and radiance components at ground and sensor levels. Three input data are required to solve this inverse problem: atmospheric parameters (aerosol type and visibility), 3D digital vector models of the scene, and spectral at-sensor calibrated images of the scene. The validation of (CARE is checked through the CAPITOUL field campaign, carried out over Toulouse (France). Measurements were performed using two airborne Pelican image systems consisting of 8 high spatial (20 cm) and spectral (30 nm) resolution cameras. Results are presented over typical urban structures in Toulouse center. ICARE performance is expected to be better than 0.04 in the reflectance retrieval, even in shadowed areas. To quantify the gain brought by (CARE, a comparison between classification based on radiance and retrieved reflectance images, obtained by inversion, is introduced. The results show that the classification is improved from 54% for a flat ground assumption to 74% using ICARE.