An Automated Digital Elevation Extraction System
3 Digital System – Advantages
Proposed here is a digital system for automated elevation extraction from aerial images. The obtained photographs are assumed to be scanned by high resolution scanner into digital images. A digital system offers a variety of advantages over conventional processing using photographs. Images taken under poor lighting, haze, sun illumination angle effects and other similar problems can be easily corrected in a digital system. Integration of images taken at different times, cameras, positions are easily achieved in he digital domain for purposes of comparison and utilization of different sensors to maximum advantage. Archival, reproduction of results, consistency in output product quality, quality control, throughput, information management and updating records are easily done in the digital domain. Automation also becomes an easier task thus minimizing valuable human effort. A extensive wealth of knowledge is available in processing digital images which can be put to use.
4. Initial Registration
The digitalized images are first enhanced if needed. Identifying the fiducially in the images space and making a correspondence with the camera calibration information yields a user coordinate system (UCS). Relates the sample/line in the digital image to the film coordinates. This step also yields the principal point in the two images, the origin in the film coordinate space. Features are displaced in the X dimension between the two images due to elevations of images features. They can also be displaced in the Y dimension primarily because the photographs may not be truly vertical. Differences in the flying height for the two images can also cause Y displacements. This situation is generally rare. In the stereo. Similarly the same can be done in the digital domain for viewing the images in anaglyphic stereo or through the stereo viewer. The Y-displacement of features can be very significant even for very small tilt angles. Though the automated correlation process can account for this problem to a certain extent, the Y-parallax is better removed (may not be completely) either by aligning the base axis of each image with the X-axis or more precisely by suing the camera model position and orientation information to compensate for the platform distortions. The correction of the left and right images to true vertical images can also be dept in just a mathematical form and no vertical images are created as such. For ease of discussions below let us assume that two new images are created as such. For ease of discussions below let us assume that two new images are created to form a stereo pair. The new vertical images will hence be called the right and left images.
5. Automated Correlation – An Introduction
When the operator uses a stereo plotter to look at a pair of photographs and keep a moving floating mark on the ground he/she is performing a manual matching to obtain corresponding pair of points in the stereo pair. The matching however is done with a lot of intelligence and an understanding of the terrain and the features. Without such a knowledge and positioning the stereo. The operator faces difficulties when the dot is moved across a desert or a forest where there are no distinct features to match. Matching becomes relatively easier in such areas when the surrounding terrain is identifiable or when there are unique features within the forest, say a forest outpost.
Mathematical matching processes have been implemented by various researchers in an attempt to replace the human matching process performed through the stereo viewer. The overlapping images are first relatively registered so that a stereo pair results. This eliminates the parallax in a direction perpendicular to the flight line ( the Y-parallax). Now the features are displaced only in the X direction ( along the flight line) as a result of their heights. Consequently epipolar lines are found at the same Y locations in both images of the pair. The relative registration can also be kept in mathematical form instead of performing a correction on the digitized images. In such a case a mathematical equation is used to locate epipolar line.
