Algorithm Research And Application Of Target Tracking

Target tracking technology is proposed in the 50's of last century, and after half a century of researching and development, it has become indispensable in modern society of high-end technology. It is playing a broad role in the military and civilian and plentiful results have been acquired in these fields. Multi-target tracking problem is the more area of concern in the field of target tracking at present, and the multi-target is a general trend in practical applications. Data association is the most important and difficult aspect in the multi-target tracking and it is the key to the final tracking results. At present, many domestic and foreign researching institutions have carried out some related research work, but to the technology widely used in the practical engineering, more theoretical research is still need in the field of target tracking. On the basis of this objective, some research works of the area have been studied in this paper, and mainly described as follows:1. In this paper, data association algorithm which is the core of multi-target tracking is analyzed and discussed, some typical data association techniques and the resulting filtering algorithm are introduced, and the basic filtering algorithm was simulated and comparative analysis, such as the Kalman filter algorithm and the probabilistic data association algorithm. Kalman filter has better filtering performance and less time overhead for the linear Gaussian and only one target problem, while the probabilistic data association filter algorithm have large computational load, but better tracking performance. However, the two filtering algorithms are less than ideal results while the multi-target is in clutter.2. Joint Probabilistic Data Association Filter shown very strong tracking performance in multi-target tracking, but it need to calculate the feasible events, and especially when the number of targets are large and in clutter, the number of feasible events increased exponentially, so that the combinatorial calculation too vast to apply in practical projects, In order to reduce the computational load of joint data probabilistic association, a modified algorithm of JPDA is proposed, which is inspired by the principle of Bound and Branch algorithm. The calculation of scanning the echoes which will not fall into the target track gate is omitted during the process of dividing confirmed matrix into feasible matrices, As a result, the time complexity is reduced significantly while the tracking accuracy and feasible events are not changed, making it favorable to realize real-time tracking.3. Finally, in a multi-target tracking, there is a problem with the number of targets unknown or varying with time and some of the target number of tracking algorithms are carried out under defined conditions in the previous article. For this problem, we propose the probability hypothesis density filtering algorithm which based on the Gaussian model that take the target set as a random set, thus data association is avoided, The simulation results show that the probability hypothesis density filter algorithm can track the target of state and the number of target stably, compare with the joint probabilistic data association algorithm, shows the advantages and disadvantages and applicability of the two algorithms.This paper also describes the latest research on target tracking developments and trends, and further the work of the next prospect.