Thermal infrared cameras have the capability to operate day and nightand to display moving objects in image sequences sampled with videoframe rate or better. However, compared to standard video in thevisual domain, these cameras have disadvantages concerning the geometricresolution. In this paper, a method for the detection of moving objectsin airborne thermal videos is presented. For the frame to frame registration,points of interest are extracted and tracked through the sequence.Then, the motion of the sensor is estimated using projective planarhomographies as the transformation model. A model for moving objectsis used to reduce the false positives. Results on four differentdatasets are shown. The introduction of the model results in a reductionof the rate of false positives by 98% compared to the method withoutthe model. The restriction of the image motion to a transformationby a planar homography highly reduces the computational effort andallows an initial estimation for sensor movement.     «

Thermal infrared cameras have the capability to operate day and nightand to display moving objects in image sequences sampled with videoframe rate or better. However, compared to standard video in thevisual domain, these cameras have disadvantages concerning the geometricresolution. In this paper, a method for the detection of moving objectsin airborne thermal videos is presented. For the frame to frame registration,points of interest are extracted and tracked through the sequence.Then, the moti...     »