The Study On Data Fusion Of Target Recognition In Combat Command Decision-making Supporting System

Combat command decision-making system is the main supporting platform to achieve the command and control in the modern warfare, Multi-sensor fusion is the prerequisite and basis to provide the command and control and helps making the decision. The decision-maker relies highly on the target fusion recognition degree to analyze, to decide, to balance the battle situation. Presently, the target fusion recognition of the decision making level has no way to make the effective uncertainty measure on the decision-making system. This paper deeply studies the pivotal problems of target fusion recognition based on rough set theory and other data mining methods. The main work is as following:1. The uncertainty measure is the key point of the research on rough set theory, based on the approximation quality y, the uncertainty measure of the decision-making system can not be comprehensively reflected. Regarding this aspect, this paper builds a measure criterion on the variable precision conditional entropy of the scoped boundary. Theoretical analysis states clearly that in case of weighing the consistency of decision system, this measure criterion and the conditional entropy has something in common on recognition, and this measure criterion is also equivalent to the relative regular domain. Experiment result indicates that compared with other measure standard, this one has the advantage of simple calculation, high accuracy, anti-noise data while carrying out the discretization processing and algorithm for attribute reduction.2. The global approach of the discretization process is difficult to keep the consistency of the system, the local approach will result in many cut points. Regarding this, this paper puts forward the discrete method of the overall continuous attributes which is based on the importance of attributes. Grouping objectes according to consistency degree to select candidate cut points can reduce the number of cut points. Gradually performing the discrete process according to attribute importance, the consistency of the system can be maintained. This algorithm concepts the importance of the cut point based on the conditional entropy of the scoped boundary, eliminate cut point with zero importance, so as to improve the running efficiency of the decision making system, reduce the scales of the decision table. Experiment result shows that this algorithm has low time complexity, less discrete cut point creation.3. The attribute reduction method based on algebra view is not applicable to the inconsistency decision system. Regarding this, this paper combines algebra view and information view to put forward the attribute reduction algorithm on the improved discernible matrix and the information entropy. The improved discernible matrix could get the core attribute in the inconsistency decision system. The attribute reduction algorithm takes the variable precision conditional entropy of the boundary as the uncertainty measure, and make the core attribute as the iterative initial value, and take the attribute importance of information entropy as guided lines. The result indicates that this algorithm provides less reduction number and formula number, it has comparatively higher recognition accuracy, for those data with noise type it can also achieve better effect.4. Combine the rough set theory and other artificial intelligence technology, the author puts forward the target fusion model of the plot-track association and target type recognition. In case of the difficult process on multi target number and close target position by plot-track association, by building the similarity equation of observation data, this paper works out the plot-track association algorithm according to the hierachical and density clustering. The experiment result states that this algorithm improve the accuracy of the plot-track association, decrease the time complexity and easily to achieve the expected effect. As for the targets which do not deduce the type through rules, this paper proposes a target type recognition algorithm, by using the rough set similarity set to give the concept of the relative function of DS evidence theory. Case studies that compared with the traditional recognition method, this algorithm owns higher calculation efficiency and recognition accuracy.