Track initiation algorithm based on Hough transform and clustering

Hough transform based track initiation method has been proposed as an effective means for low signal-to-noise ratio target detection in dense clutter environment. However, it is difficult to choose a proper threshold for peak detection and there exists a problem that multiple tracks might be initialized by only one target. A Mean Shift Multi-Scale Clustering based Hough transform track initiation (MSMSC-HT) algorithm was proposed, which first obtains the rough candidate tracks with a lower threshold, then filters the histograms in a multi-scale way according to the theory of multi-scale space, works out the local maximum of the filtering function under different scale via Mean Shift algorithm, and finally determines the track number and parameters adaptively through the life time of the local maximum number and the drift speed of the position. Simulation results show the efficacy of the proposed algorithm.