Title :
A new heuristic for multisensor PHD filter
Author :
Bozdogan, Ali Onder ; Efe, M. ; Streit, Roy
Author_Institution :
Electr. & Electron. Eng. Dept., Ankara Univ., Golbas, Turkey
Abstract :
A new approximation to carry out the measurement update step for the multisensor PHD filter is introduced. This method approximates Bayes posterior process as the union of detected and missed targets. The detected target pdfs are extracted around approximately maximum likelihood estimate of target states. Reduced Palm intensity function averaged among sensors is used to approximate the intensity due to the missed targets. For low probability of detection and/or higher false alarm rates, compared to the iterated corrector heuristic the new approximation is shown to provide sharper peaks around target states.
Keywords :
Bayes methods; approximation theory; maximum likelihood estimation; object detection; sensor fusion; Bayes posterior process; detected target pdf; detection probability; false alarm rates; maximum likelihood estimate; measurement update step; missed target; multisensor PHD filter; reduced palm intensity function; target states; Approximation algorithms; Approximation methods; Joints; Maximum likelihood estimation; Object detection; Sensors; Target tracking; PHD filter; Poisson point process; conditional intensity; multisensor tracking; reduced Palm process;
Conference_Titel :
Information Fusion (FUSION), 2014 17th International Conference on
Conference_Location :
Salamanca