Imagery simulation of space-borne multispectral sensors via reflectance characteristic modeling of Earth surface

In this work, an imagery simulation scheme is proposed to generate high-fidelity images for space-borne multispectral sensors. For this purpose, clear-sky remote sensing images of a specified region are segmented based on their spectral reflectance features using peak density research. For each class within the segmented images, the Rahman model is used to characterize the bidirectional reflectance distribution function (BRDF). Additionally, ground elevation data is incorporated to calculate the local incident angle and viewing angle for each pixel based on the observational geometry. This process establishes a reflectance characteristic model for the specified region. To simulate images acquired at different observation times and sensor viewing angles, the projected geographical position of each pixel is determined to identify its corresponding ground class and retrieve the associated BRDF parameters. To address the inevitable presence of clouds in remote sensing images, measurements from several non-target bands within the target observational scene are utilized during image classification to distinguish clear-sky pixels from cloudy pixels. For the target band, the radiance of clear-sky pixels is simulated using the reflectance characteristic model, while the radiance of cloudy pixels is calculated based on the reflectivity ratio between the target and non-target bands. The proposed scheme was applied to simulate MODIS images for shortwave infrared bands to demonstrate its effectiveness. Furthermore, the structural similarity (SSIM) between the simulated images and the actual measurements was quantitatively analyzed to validate the simulation accuracy.