A Maximal Fuzzy Entropy Based Gaussian Clustering Algorithm for Tracking Dim Moving Point Targets in Image Sequences

After targetspsila original states were estimated by multi-frame detection method, the tracking windows in which each target may be occur were used to lower the computational load. Then all the observational data could be positioned in a observational matrix, and we used a maximal-entropy Gaussian fuzzy clustering method to get the membership for each measurements to replace associated probability in traditional PDA filter, then the targetspsila following states were estimated by Kalman filter. This paper gives a new weight distribution scheme for deciding the uncertainty of measurements, and defines maximum effective distance based on difference factor to eliminate non-effective observational data. This method avoids tracking false targets or losing targets when targets are crowded in traditional target-tracking methods, and reduces greatly the computation load and has guaranteed the tracking accuracy.