Algorithm Research And Application Of Target Tracking

Target tracking is a very active topic in computer vision, image processing and pattern recognition. The core part of target tracking is filtering. With the recent development of science and technology, hardware, algorithms and computer performance and other aspects of great progress, the technology of target tracking develop rapidly. Thus, in national defense, aviation and aerospace as well as other areas of the national economy, all kinds of new technologies have been applied to target tracking in order to adapt to more complex environments and applications.This research work is mainly carried out based on the distributed target tracking. First, this paper describes the development of target tracking and target tracking algorithms overview of the main methods. Secondly, multi-platform multi-target tracking, interacting multiple model algorithm-based (IMM) algorithm to improve the basic principle and ideology, combined with joint probabilistic data association algorithm (JPDA) to achieve a single platform for multi-target tracking algorithm, and of further Extended to distributed multi-target tracking algorithm (DIMM_JPDA). Finally, in tracking multiple maneuvering targets, this algorithm will be integrated to the current statistical model IMM algorithm, the algorithm of maneuvering targets with good results. Mainly modified as follows:(1) In the maneuvering targets. Lang Hong proposed distributed based on interacting multiple model algorithm (DIMM_CA / CV), the algorithm is mainly used in single objective environment, in a multi-target environment do not apply. And the algorithm using constant acceleration model (CA) and the uniform model (CV) relative to the interaction of larger error in the motor case. With the current statistical model of the algorithm (CS) presented a new algorithm (DIMM_CS / CV). The new algorithm in tracking moving targets, moving targets especially strong with good results, the simulation results and DIMM_CA / CV algorithm comparison and analysis of its performance advantages.(2) In the multi-target. A new model for multiple target tracking algorithm, called DIMM_JPDA algorithm. Will be based on the JPDA multi-target tracking algorithm fusion IMM algorithm used to obtain IMM_JPDA algorithm. Then use the data fusion method, to extend it into a distributed multi-target tracking algorithm (DIMM_JPDA). Since the introduction of the JPDA algorithm model, making the computation is large DIMM_JPDA, not suitable for application in practice.(3) In data fusion, the covariance matrix must be reversible. However, the actual program is running, there will be irreversible situation. This paper introduces the characteristic value of the perturbation method to solve the problem. Experiments show that after a perturbation method, the matrix becomes irreversible non-singular matrix.(4) DIMM_CS / CV algorithm is a good solution to the target mobile, but do not apply for multiple targets. DIMM_JPDA algorithm can solve the multiple target tracking, but are too computationally intensive. Therefore, the authors DIMM_CS / CV algorithm and DIMM_JPDA algorithm are combined to solve the target maneuvering and the application of multi-target case.