Automatic target detection and tracking in forward-looking infrared (FLIR) scenes is a difficult task, due to the high variability of target types and background clutter and the different manifestations that these can take in terms of changing thermodynamic states and atmospheric conditions.

Members of the center are using morphological operators in order to develop fast and robust approaches to detecting and tracking targets in FLIR image sequences. Morphological connected operators are used to extract the targets of interest and remove undesirable clutter. These operators are solely based on general size, connectivity and motion criteria, using spatial intraframe and temporal interframe information. This approach avoids the variability issue in FLIR scenes, since it does not require any target modeling, and is implemented in two steps. The first step involves intraframe processing of the sequence, based on size and relative position criteria, whereas the second step consists of interframe processing that employs a simple motion criterion based on spatio-temporal connectivity. During the second step, only a few frames must be considered at a time, making the method suitable for time progressive coding and transmission schemes. The .rst row of the .gure above depicts three consecutive frames of a FLIR image sequence. The second row depicts the corresponding results obtained after intraframe processing, whereas the third row depicts the results obtained after interframe processing. Notice that interframe processing is essential for removing false alarms at the output of the intraframe processing step.