Square root Gaussian mixture PHD filter for multi-target bearings only tracking

Bearings-only tracking is a challenging estimation problem due to the variable observability of the underlying targets. In the presence of false alarms and missed detections, the difficulty of the estimation problem is further compounded by the presence of ghost targets. This paper presents a solution to the bearings only tracking problem based on the theory of random finite sets or Finite Sets Statistics. We adopt the Gaussian-Mixture Probability Hypothesis Density filter as a basis for performing multi-sensor multi-target tracking. A corresponding square root implementation is derived to ensure numerical stability of the filter and applied to a bearings only scenario. The proposed solution is a simple, computationally inexpensive and numerically stable method for fusing multi-sensor bearings information.