A Scene-Analysis Approach To Remote Sensing
by Tenenbaum, Jay M., Fischler, Martin A., Wolf, Helen C.
Technical Note 173
Institution: AI Center, SRI International
Address: 333 Ravenswood Ave., Menlo Park, CA 94025
A scene-analysis approach to interpretation of remotely sensed imagery is described, with emphasis on applications involving continuous monitoring of predetermined ground sites. A key concept is the use of knowledge contained in various kinds of maps to guide the extraction of relevant information from the image. Geometric correspondence between a sensed image and a symbolic map is established in an initial stage of processing by adjusting parameters of a sensor model so that image features predicted from the optimally match corresponding features extracted from the sensed image. Information in the map is then used to constrain where to look in an image, what to look for, and how to interpret what is seen. For simple monitoring tasks involving multispectral classification, these constraints can significantly reduce computation, simplify interpretation, and improve the utility of the resulting information. Moreover, previously intractable tasks requiring spatial and textural analysis may become straightforward in the context established by the map knowledge. Three such tasks are demonstrated: monitoring the volume of water in a reservoir, monitoring the number of boxcars in a rail yard, and monitoring the number of ships in a harbor. The conceptual approach of map-guided image analysis is described in sufficient generality to suggest numerous other applications. Details of the map-image correspondence procedure and a general technique for locating boundaries to subpixel accuracy using map knowledge are described in appendices.
|Fischler, Martin A||Principal Scientist|
|Tenenbaum, J. M.||Alumnus|
|Wolf, Helen Chan||Alumnus|