Abstract:

A new method for small target testing is brought forward aiming at the representative characteristics of small targets in infrared images. In order to realize target detecting, the minimal normal curvature is obtained in use of differential coefficient traits of small targets, and taking this curvature as a threshold, the candidate object of small targets is acquired. Aiming at the target disappears because of background noise or target being sheltered under complex circumstances, the tracking technology for dynamic small targets in serial infrared images is researched, and a realtime tracking algorithm based on probabilistic data association filter (PDAF) and the linear prediction technology is designed so as to improve the stability and precision of target tracking. Finally, the simulation with memorized real time image sequence shows that this algorithm is able to track targets with the signaltonoise ratio no less than 2 and velocity 1 frame per pixel, which also proves the validity and realtime quality of this algorithm.