Linear- and Repetitive Feature Detection Within Remotely Sensed Imagery
U.S. Army Engineer Research and Development Center (ERDC) Hanover United States
Pagination or Media Count:
The United States Army has a variety of applications for identifying features of interest within remote imagery. Whether it is characterizing a landscape while planning operations or trying to find particular installations in an urban setting, the Army can glean a significant amount of information from imagery data. This study investigates methods that can detect linear and repetitive features contained in remotely sensed images that are in panchromatic or true-color formats. Image-processing techniques, including Hough transforms, machine learning, and template matching, are capable of detecting different kinds of features within images. However, the success of these methods depends on effectively preprocessing image data, which has proven difficult and intensive for certain images. In many cases, the amount of user interaction needed to produce useful results exceeds the amount of labor needed to manually inspect individual images. At their current state, these methods provide useful tools to help analysts detect features but do not replace their expertise. This report summarizes several techniques for preprocessing image data and then detecting linear and repetitive features in that data.