Abstract: In order to realize multi-target track initiation of infrared fisheye warning system, the technology of multi-target track initiation was studied for infrared fisheye warning system. The target moving characteristics on the image surface were analyzed, based on which an algorithm of multi-target track initiation proper to the infrared fisheye warning system was proposed. Using 3/4 logic of the logic method, the maximum velocity constraint method was firstly used to remove the uncorrelated measuring points in each sampling period so that the calculation amount of the Hough transformation could be reduced. The Hough transform method was used to generate initial target tracks, and the uniform velocity criterion was used to select the right tracks. The algorithm was verified with simulated targets. It is shown that, compared with the standard Hough transform and the logic method, the track initiation performance is obviously improved. For 450 clutters, the false alarm probability of track initiation is only 14%; the track initiation consuming time for continuous points is 3 frames while that for the non-continuous points is 4 frames. Further more this algorithm has less running time under the dense clutter environment.