Translator Disclaimer
26 September 1989 A Pattern Recognition Approach To Image Tracking In Cluttered Sequences
Author Affiliations +
The use of an infrared imager in an air-to-ground tracking and guidance application is considered. The application requires tracking certain small distinct hot or cold spots in the presence of other potentially confusing objects. The target spot must be distinguished from the surrounding clutter. A tracking method based on pattern recognition techniques is used. Several approach sequences of IR imagery taken during captive flight were recorded digitally and used for this development. For each approach sequence, the tracker was initialized on a target spot. During each image frame, several feature measurements were made of the target spot. Confusing objects in the tracker search area were also identified, and the same features were measured for each detected clutter object. The target and nontarget feature measurements from all frames of all sequences were used to formulate a two-class pattern recognition problem. The classes were found to be well separated by both linear and quadratic discriminants. When the derived discriminants were used to classify the target spot from nontarget objects, high correct classification probabilities and small false alarm and miss probabilities were obtained. This resulted in more accurate tracking performance, even in scenes containing a high density of confusing clutter objects.
© (1989) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
David Shrode, Don Van Rheeden, Joel McWilliams, and Anthony Reid "A Pattern Recognition Approach To Image Tracking In Cluttered Sequences", Proc. SPIE 1111, Acquisition, Tracking, and Pointing III, (26 September 1989);

Back to Top