A novel multi-human location method for distributed binary pyroelectric infrared sensor tracking system: Region partition using PNN and bearing-crossing location

This paper proposes a novel multi-human location method for distributed binary pyroelectric infrared sensor tracking system based on region partition using probabilistic neural network and bearing-crossing location. The detection space of system is divided into many sub-regions and encoded uniformly. The human region is located by an integrated neural network classifier, which is developed based on the probabilistic neural network ensembles and the Bagging algorithm. The location of a human target can be achieved by first determining a coarse location by this classifier and then a fine location using our previous bearing-crossing location method. Simulation and experimental results have shown that the human region can be judged rapidly and the false detection points of multi-human location can be eliminated effectively. Compared with the bearing-crossing location method, the novel method has significantly improved the locating and tracking accuracy of multiple human targets in infrared sensor tracking system.