|
Background information fusion target recognition technology for different sensors is limited by the working conditions, the same goal of a single sensor feature information and greater uncertainty as well as a single sensor recognition performance lower, explore and more information source under the conditions of an effective method to obtain useful information and extract knowledge and implementation techniques, research and discuss how to effectively use multi-source information to improve target recognition rate, the introduction of a rough set theory, rough set and neural network combining technology, focusing on information fusion in information extraction, knowledge representation, attribute reduction rules modeling, incremental learning, and research based BP and RBF network, rough neural network system. Around these issues, this paper carried out the following work: research knowledge to simplify and knowledge representation system based on rough set theory, including the basic concepts of rough set attribute reduction, rule generation and simplified, which focus on the reduction of various properties algorithm, taking into account the real-time image processing, the paper presents a quick attribute reduction algorithm to overcome the algorithm time complexity of large drawback. Problems encountered in the multi-sensor fusion, and the lack of existing technology, the use of rough set theory to create a data fusion target recognition rules model, object model and a priori knowledge of the case, based on the existing data to build fusion The object model rules, the target recognition rules Knowledge Base, and the introduction of rules of credibility and rules support the concept of assessment of classification rules. Target recognition process for image processing, real-time requirements, the image data, we propose the introduction of incremental learning in rough set theory to establish the model of image target recognition rules, accelerated algorithm processing, improve the efficiency of operations. Rough set theory and neural network to study the relationship between rough sets and neural networks, the proposed neural network learning mechanism is introduced into rough set system; rough set and decision attribute structure neural network architecture, research-based rough neural network algorithm for BP network and BP network's shortcomings and limitations, research rough neural network algorithm based on RBF network. Blend of rough sets and neural networks, good use of their respective advantages, and also to make up for their shortcomings, and improve the efficiency of the algorithm running, improve recognition rate and reduce the false alarm rate.
|