Underwater multi-target tracking using improved multi-sensor multi-Bernoulli filter

Aiming at the multi-sensor multi-target tracking problem underwater,an improved method based on multisensor multi-Bernoulli filter is proposed.There are two improvements in this method.Firstly,in order to obtain the new target state under the situation of multiple targets going in and out randomly,a target generation method based on node-paired and combined random observation is proposed.This method divides the observed nodes into pairs and performs inter-pair combination.The observation pair combination is randomly selected with the same probability to generate new target information.Secondly,in order to get continuous and stable trace,a trace generation method based on statistical binary threshold is proposed.This method accumulates the number of detections exceeding the first threshold,and compares the result with the second threshold to control the target track output.The simulation results show that:(1) The new target generation method based on node-paired and combined random observation can get new target information correctly.The algorithm complexity is proportional to the sensors number and less than the global cross-target generation method;(2) the tracking method based ot statistical binary thresholds trace generation can output a stable trace in real time,and its Optimal Sub-Pattern Assignment(OSPA) distance is reduced about 20%compared with the tracking method that do not use this trace generation.