Research On Group Target Information Correlation Techniques

Information correlation technique is the precondition and key for multi-source information fusion. It groups the information by description object and the information of each group is from one and the same description object. It lays a foundation for the subsequent fusion process.Taking the maritime surveillance application based on satellite reconnaissance as its background, the dissertation studies some basic problems in group target information correlation and group target information correlation algorithms for the circumstances when the group targets'features are steady and dynamic respectively, including the framework and the model for group target information correlation, the algorithms for the group member information matching under the steady circumstances, the algorithms for group target feature extraction and correlation decision under the dynamic circumstances, etc al. The research results of the dissertation are summarized as follows:(1) The frameworks of group target information correlation are established for the steady and dynamic circumstances respectively. The mathematical model for the key part of group target information correlation under the steady circumstances, i.e. group member information matching problem, is established. And the models for the calculation of group target information correlation measure and the correlation decision under the dynamic circumstances are also given.(2) The mathematical model for group member's position information is established. Considering that group member's position information is usually influenced by the factors such as observation noise, missing, partially overlapping observation areas, different resolution of sensors and so on, a new effective method for feature description invariant to coordinate transformation called relative Cartesian coordinate description is proposed Taking the relative Cartesian coordinate description as its base, a position information matching measure based on coordinate registration is given which overcomes the influence of,.the aforementioned factors.(3) The mathematical model for group member's attribute information is established. A new conflict measure is defined and its performance superiority over the traditional one is proved theoretically. Based on the new conflict measure the matching measure of the group member's attribute information is given. Combining with the new conflict measure a new evidence combination rule is proposed and its performance superiority over the traditional one is proved theoretically. Based on the new evidence combination rule the matching measure which synthesizes the position and attribute information of a group member is obtained.(4) Under the circumstances when the group target's position information is globally rigid, aiming at the problem that the 2D assignment algorithms is poorly applicable to missing, a new matching decision idea, called matching decision based on the truncation of information sequence, is proposed. Considering that the many-to-many matching algorithm based on the truncation of information sequence is sensitive to the perturbation of non-matching elements, the many-to-many matching decision is achieved via clustering. And the selection of matching clusters'number and centers are resolved. Aiming at the problem that the many-to-many matching algorithm based on the truncation of information sequence is sensitive to the perturbation of matching elements, the many-to-many matching decision problem is modeled as a process of perturbation correction to the matching measure matrix and the detailed algorithm design is given.(5) Aiming at the circumstances when the group target's position information is locally rigid, the concept of local subset of group target information is established. Based on this concept two composite matching decision methods which unite local matching decision and global validation are proposed. They are local matching decision and aggregate global validation, local matching decision and incremental global validation. The effectiveness and applicability of the two composite matching methods are tested and analyzed through simulation experiments.(6) The group target feature extraction and information correlation decision problem under dynamic circumstances are investigated. A general group target formation extraction method and a general group target composition extraction method are proposed respectively. And a group target information correlation decision algorithm, called correlation threshold parameterization– two-way constraint method, is given which is applicable to the dynamic circumstances.