True orthoimage production for forested areas from large-scale aerial photographs

Abstract

An orthophoto is a fundamental information source in forestry. Orthophoto production is usually based on digital elevation models (DEM); however, the elevations of 3D objects such as buildings and trees, which are visible in large-scale photos, are not included in DEMs. Recently initiated "true orthophoto" generation has been dedicated to urban scenes, removing the displacement caused by both the terrain and buildings. This paper reports an effort on reducing occlusion and distortion caused by trees in true orthoimage generation for forested areas. Canopy surface models (CSM) depicting both the terrain and tree surfaces were used in forest orthoimage production. Z-buffer techniques were incorporated into indirect orthoimage production to remove occlusion. An angle-based strategy was applied to composite multi-ocular images. The proposed techniques were applied in the generation of high-resolution orthoimages of a redwood stand with 1:2,400-scale aerial photos.