Change Detection and Object Tracking in IR Surveillance Video

The appeal of change detection methods for surveillance systems arises from the need to monitor perturbations in the sensorpsilas field of view. Applications of change detection include satellite imagery, digital video compression algorithms, medical diagnosis, damage monitoring, and object tracking. The algorithm developed in this study is capable of extracting features from static background IR (infra-red) surveillance video by a spatio-temporal change detection frame work. Those features are then tracked by application of a similarity metric. The system is trained by observing the background for an extended period of time. The constructed background model helps the algorithm to detect non-stationary objects and extract their time-varying features such as motion and shape deformation. Spatial and spectral characteristics of these dynamic features enable the algorithm to identify targets of known nature and isolate them from the surrounding background cluttered by various objects and foreground.