Please use this identifier to cite or link to this item: https://hdl.handle.net/11681/4842
Title: Knowledge-based image analysis
Authors: LNK Corporation.
Stockman, George C., 1943-
Lambird, Barbara A.
Lavine, David.
Kanal, Laveen N.
Keywords: Registration
Image matching
Map-guided cartography
Feature extraction
Verification
Region classification
Pattern recognition
Image analysis
Issue Date: Apr-1981
Publisher: U.S. Army Engineer Topographic Laboratories.
Engineer Research and Development Center (U.S.)
Series/Report no.: ETL ; 0258.
Description: Contract report
Abstract: The work reported was directed toward employing a priori knowledge in the automatic analysis of aerial imagery. Major objectives of the research were directed toward (1) map-guided registration, (2) verification of geographic data bases extracted from imagery, (3) enrichment of geographic data bases, and (4) automatic terrain feature extraction using multiple sources of knowledge and multi-level decision making. The key component in all of the work was the matching of existing iconic structure in a geographic data base (GDB) with detected image structure. By using iconic knowledge, the image interpretation paradigm becomes a three step process. First, some primitive features of the imagery must be recognized without any area-specific knowledge. Second, the imagery is aligned or registered with the knowledge base by drawing correspondences between the image features and their iconic analogues in the GDB. The matching is formalized by derivation of a transformation which maps points (x,y) of the image to points (u,v) in GDB coordinates. The final step of the process is the analysis of those parts of the image which were not successfully interpreted in steps 1 and 2. This implies a top-down search for image structures which correspond to features in the GDB. Section 2 of the report treats primitive extraction. The emphasis is currently on lineal, point, and region features only. A method for automatically inferring a rotation and translation transforming image to map is given in Section 3. Classification of registered regions is discussed in Section 4. Verification of lineal GDB features in gray-scale imagery is introduced in Section 5.
URI: http://hdl.handle.net/11681/4842
Appears in Collections:Documents

Files in This Item:
File Description SizeFormat 
ETL-0258.pdf6.85 MBAdobe PDFThumbnail
View/Open