Research On The Algorithm And Application Of Decision Fusion For Multi-source Heterogeneous Sensor

Decision level information fusion is a kind of high-level fusion technology that can provide the basis for judging and support. Based on certain rules, it can extract all sorts of useful information after low level fusion and obtain the final inference on the target identity through reasonable analysis and appropriate algorithms according to the specific target. Situation assessment and threat assessment are the two crucial parts in decision level information fusion technology. Among them, situation assessment is a multilayer view that can reflect the real situation of the environment. By analyzing the relationship between the target entity, target groups with similar behavior and purpose are aggregated and relevant situation analysis and prediction are made integrated with various environmental factors. As an important factor in command decision information fusion, threat assessment is also a key link in information fusion. By the situation assessment processing, results of earlier conclusions are deeply analyzed and enemy threat degree of our protected targets are evaluated combining all factors.In this paper, the two key technologies of decision-making information fusion in target classification and target threat estimation are discussed, spetial attention are paid to the algorithm and application field. The thesis mainly includes the following contents:1. Based on assessment theory, the main functions of situation assessment technology is first analyzed in detail, several commonly used algorithms in the field of fusion are discussed and compared. Focusing on the key technology of object clustering in situation assessment, several commonly used fuzzy clustering algorithms based on objective function are comparatively analyzed.2. Intelligent algorithm searching has good characteristics, such as strong flexibility and simple structure. Aiming at the problems faced by the initialization of fuzzy clustering algorithm, linear particle swarm optimization possibility fuzzy Cmeans algorithm is proposed and its optimization performances are evaluated by simulation data of the standard database.3. Robust adaptive particle swarm optimization for clustering analysis based on steepest descent method is proposed, which is mainly used to solve the parameters setting problem of particle swarm optimization. This method enhances the searching ability and improves the algorithm’s searching speed and adaptability to a large extent. The search ability and the effectiveness of the robust clustering algorithm are proved by comparison of simulation results.4. Intelligent BP network algorithm based on immune mechanism is put forward. First of all, immune mechanism is introduced in the particle swarm optimization. By adjusting the individual concentration, diversity of the population is enhanced and premature or local optimal solution is prevented. Secondly, by using the immune particle swarm optimization algorithm to obtain the optimal weights and threshold, initial network is constructed and threat degree estimation of the target are accomplished through network training. Finally, through simulation and contrastive analysis, the effectiveness and performance of the algorithm are verified considering accuracy, convergence, stability and practical application.