Parallel Bin-Occupancy Filter for Image Observation

We present a recursive track-before-detect (TBD) algorithm based on the bin-occupancy filter for multi-target tracking from image observations by asking a different question: "Is there a target within a given pixel?" Analytic posterior probability distribution of the bin-state is derived for image measurements under the non-overlapping assumption. Coupled with the intrinsic grid characteristic of its measurement-space, the proposed algorithm can be regarded as a fully discrete version of the traditional probability hypothesis density (PHD) filter for TBD via quantized state-space modeling. Parallel grid implementation is also discussed in order to meet the requirement for realtime processing applications. Simulation results are presented to demonstrate the effectiveness of the proposed algorithm for tracking targets with low signal-to- noise ratio (SNR).