Research On Target Recognition Technologies Of Battlefield Based On Neutral Networks

Along with the information time arrival,the information gain is even more important in the war status.The sensor already massively used in the battlefield information gain,carried on the accurate prompt recognition,through the sensor to the battlefield goal,increasingly became the key of victory.However as a result of the electronic countermeasure technology massive applications,causes the battlefield environment to be day by day complex,is under the many kinds of factor influence,under the research battlefield special environment the multi-sensors goal recognition question,uses the multi-sensors realization goal recognition,greater degree collection and processing goal information,thus the enhancement recognition system accuracy and the reliability has the vital significance.According to problems of multi-sensor target recognition in a complex interference environment,the thesis presents the theoretical research on them.Firstly,the thesis analyses the shortcomings that exist in traditional attribute fusion architectures,and brings forward Decision Fusion Theory based on entire information fusion,which provides a scheme that resolves problems of multi-sensor target recognition in a complex interference environment. Secondly,according as traits recognition in a complex interference environment,several correlative technologies are researched:BP neutral networks is used for classification,which gets over disadvantages of complex interference environment:lack of knowledge of target attributes and little samples which fluctuate vary irregularly; through analyzing the influence of the sensor state, sensor performance and environmental aspect on the observations of targets, this thesis employs ANFIS confidence estimator that is built up by neural network and fuzzy reasoning. ANFIS confidence estimator obtains the credibility of each sensor by an adaptive synthetical processing of information of sensor state, sensor performance and environmental aspect. Finally, the thesis presents a system model of mufti-sensor target recognition in a complex interference environment, which is realized by the above-mentioned algorithms. This system model is simulated in hypothetic environment by computer.