Tracking CSOs Using PHD Filter from Image Observations

Aiming at tracking Closely-Spaced Objects (CSOs) from image observations, a new method using the Probability Hypothesis Density (PHD) filter is proposed. To circumvent the unresolved measurements problem, a detection process is used firstly to extract the connected sets of object pixels that likely correspond to the unresolved targets of interest. Then the representative measurements are constructed to cast the CSOs tracking in the framework of PHD filter. The newly resolved targets are naturally modeled as spawned targets thus can be detected and estimated immediately by setting appropriate target spawn intensity. Gaussian mixture (GM) implementation is used for this filter, and simulations are carried out to verify the effectiveness of the proposed method.