Terrain Shadow Detection from RapidEye Images Based on Hillshade Algorithm

초록

To enable quick and ease utilization of satellite imagery, preprocessing of satellite imagery is necessary. Pixels that hinder accurate analysis should be identified and masked out. Such pixels include shadows. They disturb the irradiance from the actual earth surface and create confusion for dark objects such as water. In mountainous areas, terrain shadow occurs as a result of the terrain relief effect. It leads to accuracy in the analysis of change detection, classification, and other indices like NDVI. In this paper, we developed an automated technique to detect terrain shadow from satellite images at a 5m Ground Sampling Distance. A hillshade algorithm takes into account the sun’s elevation and azimuth angles and the surface’s slope and aspect angles to produce realistic terrain shadow effects. we demonstrated this algorithm could be applied to detect terrain shadows present in satellite imagery and to mask out shadow pixels accurately. To test the proposed method, the analysis was performed on mountainous areas using RapidEye images. Geometric and orthogonal corrections were applied to the satellite images. The sun’s elevation and azimuth angles were prepared from the metadata of the satellite images used. The surface’s slope and aspect angles were estimated from a Digital Elevation Model generated by the National Geographic Information Institute of Korea. The vectors for sun illumination direction and surface normal were then calculated. Shadow pixels were determined by calculating the angle between the two vectors and checking whether the angle exceeded 90 degrees. We then compared the predicted shadow pixels with the ground truth data. As a result, the overall accuracy of the prediction results was found to be greater than 85%. We conclude that our proposed method appears to be suitable for detecting terrain shadows. © 43rd Asian Conference on Remote Sensing, ACRS 2022.

키워드