The current bottleneck in wide area persistent surveillance missions is slow exploitation and analysis (real-time and forensic)by human analysts. We are currently developing an automated data exploitation system that can detect, track, and recognize targets and threats using computer vision. Here we present results from a newly developed target detection process. Depanding on target size, target detection can be divided in three detection classes: unresolved targets, small extended targets, and large extended targets. The Matched Filter (MF) method is currently a popular approach for unresolved target detection using IR focal plane arrays and EO (CCD) cameras and sensor detectors. The MF method is much more difficult to apply to to the extended target classes, since many different matched filters are needed to match the different target shapes and intensity profiles that can exist. The MF method does not adequately address non-fixed target shapes (e.g. walking or running human). We have developed an approach for robust target detection that can detect targets of different sizes and shapes (fixed/non-fixed) using a combination of image frame time-differencing, deep-thresholding, and target shape and size analysis with non-linear morphologial operations. Applications for gound vehicle detection under heavy urban background clutter will be presented.
Improvements in remote sensing technology for the collection of high resolution aerial LIDAR and hyperspectral data of
urban landscapes have led to increasing interest in rapid 3d scene reconstruction and environment inferencing. In recent
years algorithmic strategies fusing aerial LIDAR with hyperspectral data have been proposed to increase the overall
confidence in data segmentation by taking advantage of the unique qualities of the data for each sensor type. No
technique exists today, however, that fully automates the end-to-end process - from the initial collection of the
uncorrected data to the production of a finished, accurate and realistic urban scene. Notwithstanding, key milestones that
minimize human intervention have been made, and notable high quality suites of semi-automated tools are available
today. In this paper, an alternative strategy towards fully automated building extraction under a variety of terrain relief
conditions is presented. Advantages are discussed of multiple sensor feedback loops that update the scene at each
segmentation step starting with an initial hypothesis of each feature's classification. The merit of such a strategy from the
point of view of implementing it within a fully automated system is presented.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.