-

In this research work a shape from shading (SFS) technique which incorporates a Lambertian model is implemented for the automatic generation of a digital terrain model (DTM) using a single view aerial image. The developed algorithm is tested on both simulated and real data. The estimated accuracy of the generated DTM from the simulated data, which has a bilinear surface, is about :t3 cm'. The real data is a scanned aerial photograph taken over a low texture hilly terrain. The generated DTM by the SFS technique is compared with a DTM acquired by the manual measurement of the stereo image of the same area using a photogrammetric plotter. The estimated rimes of the discrepancies between the grid nodes of the measured and the automatically generated
DTM is about :t4 meters. The unsuccessful



























reconstruction of the terrain surface for the real data is due to the fact that a simple Lambertian model does not take into account, in a perfect way,' different nondeterministic influential factors such as the terrain alb Edo variations and the random noise. The influence of the latter case was reduced by a low pass filter applied as a preprocessing stage.