Accession Number : ADA376824


Title :   Computer-Aided Recognition of Man-Made Structures in Aerial Photographs


Descriptive Note : Master's thesis


Corporate Author : NAVAL POSTGRADUATE SCHOOL MONTEREY CA


Personal Author(s) : Cardoso, Luiz A


Full Text : http://www.dtic.mil/dtic/tr/fulltext/u2/a376824.pdf


Report Date : Dec 1999


Pagination or Media Count : 161


Abstract : Aerial image acquisition systems are producing more data than can be analyzed by human experts. Most of the images produced by remote sensing satellites, including military ones, never get seen or inspected. In this work, automated detection and recognition of buildings in aerial photos is explored. Connectivity analysis is performed on graphs derived from line segment representations of the original images, obtained with the use of the Radon Transform. The model is experimentally validated using 2-meter panchromatic aerial photographs from the National Aerial Photography Program (NAPP), which provide a marginally adequate resolution for the recognition of small buildings.


Descriptors :   *PATTERN RECOGNITION , *AERIAL PHOTOGRAPHS , NEURAL NETS , AUTOMATION , TARGET RECOGNITION , THESES , MILITARY APPLICATIONS , COMPUTER VISION , TARGET DETECTION , PIXELS , REMOTE DETECTION , MULTISPECTRAL , IMAGE REGISTRATION , SATELLITE PHOTOGRAPHY , FEATURE EXTRACTION


Subject Categories : Cartography and Aerial Photography
      Cybernetics


Distribution Statement : APPROVED FOR PUBLIC RELEASE