Multiple Maneuvering Point-target Tracking Using Filter Bank in an Ir Image Sequence

Tracking of maneuvering multiple point-targets in the presence of clutter, in IR image sequences is indeed a challenging task. In real world applications, no apriori knowledge about the movement of the target is available. We propose a method to track maneuvering and nonmaneuvering targets with large motion (±20 pixels) using multiple filter bank in an IR image sequence with occlusion due to clouds and background noise. The switchover between the filters in the bank is based on a single-step decision logic and consequently, is computationally more efficient, and performance-wise more robust. We also present a nearest neighbor data association method using the state vector of the model.