An Automated Digital Elevation Extraction System
Ram Srinivasan
1. Abstract
A digital system is presented that automates the data from a pair of overlapping aerial images for use in digital mapping. The digitized images are registered using the camera middles to eliminate platform induced distortions and obtain a stereo pair. An automated process can then be initiated to correlate between corresponding points in the stereo pair at any desired regularly spaced interval. An intelligent process has been developed to overcome the difficulties in false correlations and to minimize the computational burden by adapting to the terrain. The correlation process provides the disparity, or the parallax information which is then converted to true elevation data and ground location using the absolute orientation parameters. In regions where a more dense grid of elevations are needed the operator identifies them by delineating regions where a more dense grid of elevations are needed, the operator identifies them by delineating regions-of-interest interactively in graphics mode. The automated elevation extraction process can also be used to supplement data obtained by traditional means such as stereo plotters. The extracted data is then interpolated to a higher resolution grid and edited manually or automatically digital filtering operations. Elevation data can be used for many purposes such as producing contour elevation maps, ortho rectified images and perspective scenes for simulation.
2. Background
Obtaining elevation data from multiple aerial images is an important step in many applications such as ortho image generation, map making, geographic information systems, and mission planning. Currently prevalent methods are mostly analog in nature. Flights are planned to acquire overlapping photographs covering areas of interest. Typically there is a 60% overlap between adjacent photographs. The diapositives made from photo negatives are then placed in a stereo plotter, and a human operation of the diapositives until the images are seen in stereo through the viewer. At this stage the necessary camera model, calibration and other orientation information have been obtained that will enable computation of elevation values from a knowledge of the measured parallax or disparity between
conjugate points in the two diapositives at various locations. The operator positions a floating mark or dot on the ground viewed in three dimensional space. Through optical/mechanical means the operator can move the dot up or down relative to the surface of the stereo model. The single dot essentially represents a combination of two of them, one for each photograph. When the dots appear to merge, and a single 3D dot is placed on the ground, the different in X locations of the two dots in each respective depictive represents the disparity or the parallax between the conjugate points. As the dots are moved in the vertical dimension to match the surface of the stereo model the plotter then obtains the parallax information at equally spaced intervals, converts the measured parallax values to elevation data and writes it to a file along with the corresponding X/Y ground location. Typically, elevations are obtained along vertical lines called profiles. Elevations can also be obtained as contours. The obtained elevation data or elevation model is then fitted to a regularly spaced grid. Elevation data can be obtained from the U.S. Geological Survey or the Defense Mapping Agency of the US in gridded form. Two types of data are available called the DEM (Digital Elevation Model) and DTED ( digital Terrain Elevation Data), each of them meeting different accuracies. The gridded elevation data is then used for various purposes, important among them being ortho rectified photograph production. Sometimes it sis also important to edit the elevation data, particularly for quality control of elevation data that is widely distributed.
Besides aerial photographs digital satellite stereo images such as from SPOT are being used increasingly. The usage of satellite digital data has been more prevalent in the developing countries, primarily because the developing countries are leap frogging with the technological advances taking place. Though the article is primarily centered around aerial images, the system is equally applicable for satellite images with modifications in the mathematical modeling process.
