Research On Multi-group Target Tracking Method Based On Network Flow

The multi-target tracking is a process to estimate the characteristics of multiple moving objects based on sensors,and its technology is widely used in robot sensing systems,intelligent traffic monitoring systems,radar warning and tracking and so on.With the introduction of the new combat concept of unmanned aerial vehicle(UAV)group,uncertain factors such as strong mobility,structural evolution and high density of UAV groups,coupled with a series of external uncertainties such as the increasingly complex modern battlefield environment and detection blind spots,make the existing target tracking methods unable to accurately and effectively correlate the group target and the measurement.It seriously affects the tracking effect of multiple groups of targets.Therefore,for the problem of multi-group target data association under unknown clutter information,this paper proposes a multi-group target tracking method based on network flow.The research contents are as follows:1.Since it is difficult to distinguish the target measurement from clutters under unknown clutter information,so that the measurement cannot be segmented correctly,we propose a hierarchical measurement and segmentation algorithm based on DPeak-ASCluster joint clustering.Firstly,the density peak clustering is used to calculate the local density of any shape measurement and remove the global clutters.Then the adaptive spectral clustering algorithm is used to segment the measurements.According to the Laida criterion,the clutters around the measurement are got rid of.Finally,we obtain the number of group targets at each moment and the measurement set corresponding to each category.The simulation results show that the proposed algorithm has higher measurement and segmentation accuracy than spectral clustering algorithm and DBSCAN algorithm.2.In the actual tracking process,the number of group targets is time-varying,and the split/merge of group targets is uncertain,which makes it impossible to accurately obtain the correct relationship between group targets,we propose a multi-group target tracking method based on min-cost network flow is proposed.In combination with the results of segmentation,we establish a multi-group target network flow model and design a minimum cost optimization function under multiple constraints.Then the A*search algorithm is used to extract multiple target trajectories.Finally,the optimal associated trajectories are obtained by the direction-implicit speed constraint conditions.The simulation results show that,compared with the multi-group target tracking algorithm based on random finite set theory,the proposed algorithm has stronger robustness.3.Due to the limitation of the radar monitoring range or the sudden failure of the detector,the group target measurement sets are missing.we propose a hierarchical network flow multi-group target tracking algorithm based on intermittent observation.First,according to the low-level network flow model we obtain a reliable set of tracklet segments.Then,we build a high-level Tracklet network by using the obtained set of tracklet segments.The A* search algorithm is used to obtain the correlation solution set of each tracklet segment.Finally,the extracted trajectories are optimized by unequal interval filtering to obtain multiple optimal group target trajectories.The simulation results show that,compared with the multi-group target tracking algorithm based on random finite set theory,the algorithm proposed in this paper can more accurately obtain the data association results between measurement sets and track segments,and has a better tracking effect.