| Abstract | Digital elevation models (DEMs) extracted from high-resolution stereo images (SPOT-5, EROS and IKONOS) using a three-dimensional (3-D) multi-sensor physical model developed at the Canada Centre for
Remote Sensing, Natural Resources Canada were evaluated. Firstly, the photogrammetric stereo-bundle adjustment was set-up with few accurate ground control points. DEMs were then generated using an area-based multi-scale image matching method and then
compared to 0.2-m accurate lidar elevation data. Elevation linear errors with 68% confidence level (LE68) of 6.5 m, 20 m and 6.4 m were achieved for SPOT, EROS and IKONOS, respectively. The worse results for EROS are mainly due to its asynchronous
orbit, which generate large geometric and radiometric differences between the stereo-images. When these differences are not large (such as in the middle of the stereo-pair), 10-m LE68 was achieved. Since SPOT and IKONOS DEMs were in fact a digital
terrain surface model where the elevation of land covers (trees, houses) is included, the elevation accuracy is performed depending on the land cover types. LE68 of 1-2 m were obtained for bare surfaces and lakes. However, when compared to sensor
resolution, SPOT achieved better results than IKONOS: half-pixel versus 1.5 pixels. On the other hand, LE68 of 4 m to 6.6 m were obtained depending on the forest types (deciduous, conifer, mixed or sparse) and its surface elevation. |