The Research And Applications Of Multi-sensor Data Fusion Algorithm

Multi-sensor data fusion is a multi-source data processing technology, which can combinemulti-source data from various sensors and obtain more reliable and rational data than thesingle-sensor data fusion. It has a wide use in the domain of military and civilian.In this dissertation, an overview of multi-sensor data fusion is given, the research status of datafusion is introduced, and a basic model of multi-sensor data fusion is provided, which is includingfive parts: data preprocessing, target state estimation, situation assessment, threat assessment andinfo-feedback and correction. Then some basic algorithms and typical applications of data fusion areintroduced. In this dissertation, the multi-target tracking problem is deeply studied, and the dataassocation problem is focused on. For the plot-track association problem in multi-target tracking, theFCSS (Fuzzy C-Spherical Shells) algorithm is used to solve it, and a FCSS-based multi-targettracking algorithm is proposed. Compared to many other target tracking algorithms, this algorithmdoes not need to fully estimate the targets’ current positions, but the targets’ moving distance betweenthe two adjacent moments, and then the plot-track association is processed by using FCSS algorithm,so as to update the targets’ current tracks and achieve the target tracking. Simulation resultsdemonstrate the feasibility and effectiveness of the algorithm. For the target recognition problem, thecomprehensive information of the targets’ electromagnetic radiation information, targets’ kineticsinformation and the human judge result of the targets type are considered, and the measuring data canbe divided into three strands by these three aspects. Then use gray correlation analysis method toprocess the measuring data, build three basic probability assignments (BPA), combine the BPAs byDempster’s rule, calculate the belief function, and analyse it to provide the judge result of the targets’type. Simulation results verify the feasibility and effectiveness of the alogrithm. Finally, a prototypesystem of situation awareness and threat estimation is designed and implemented. This system, whichprovides a simple simlation environment for those algorithms, mainly includes some modules:generation of target measuring data, generation of noise data, management of device parameters, datafusion, intergration and analysis of fusion results, etc.The novelty of the dissertation is that a FCSS-based multi-target tracking algorithm is proposed,and the simulation results demonstrate the feasibility and effectiveness of the algorithm, which has ahigher tracking accuracy than the FCM-based multi-target tracking algorithm.