Previous : Introduction Up : Contents Next : Point processes



Digital Elevation Models


Usually, DEMs are built from aerial images. The data used in this work has been provided by French Mapping Institute (IGN).

To built a DEM, one requires two (or more) images of a scene. Knowing the position of the cameras that have taken these images, it is possible to built an image of the height of the scene, as does the brain with the two images taken by the two eyes to built a 3D representation of the environment.

Figure a) shows two aerial images of the French town of Amiens. With images of this type, the French Mapping Institute (IGN) has built the DEM shown on figure b). On this DEM, white pixels represent high points, while black ones stand for low points.

The DEM is of medium resolution (20 cm by 20 cm in the horizontal directions, and 15 cm vertically). Figure c) shows this DEM in 3 dimensions, under several points of views.


Comments

The noise we are interressed to deal with, is not a usual image processing noise, due to the algorithms used to construct the DEM.

Mainly, it is a geometrical noise (presence of trees, cars).

Other problems come from the density of the urban area. Some buildings are not visible on one of the images due to occlusions. Thus, for this kind of buildings, it is impossible to construct a 3D representation since some information is missing. This is known as the occlusion phenomena.

Remark about the figures : presented images were obtained by cropping. By cliking on an one of these images you can get the entire image.

Figure a) Two aerial images (20cm) of the French town Amiens, provided by the French mapping Institute (IGN)

Figure b) Obtained DEM, provided by IGN.

Figure c) 3D representation of the DEM under two different point of view.



Previous : Introduction Up : Contents Next : Point processes



Back to M. Ortner's home page !