No abstract

No abstract

Case-Based Reasoning(CBR) is a recent approach to problem solving and learning. Originating in the US, the basic idea and underlying theories have spread to other continents.In this paper,A undeIWater target recognition system based on CBR is designed.A naval vessel's noise is a initial problem definition, its type is this problem solution, the feature vector of naval vessel's noise and its type is regarded as a case.Applying a stepwise approach to retrieve a best match case from previous cases, and then the best match case is used to identify the type of undeIWater target. Experiment resultes have showed that the system has better adaptability and more higher correct recognition probability.

A dynamic recognition method using support vector machine (SVM) is researched in the paper, which is made up of the SVM based on target feature and the k-nearest neighbors. The SVM based on target feature is put forward by integrating the target feature with SVM, whose role is recognition of the target feature. It searches the optimal separating hyperplane of the local space taking the target feature as center, but does not search the optimal separating hyperplane of the whole space. To show better importance of each sample to the target feature, a method is put forward that the penalty function C i is measured. And the dynamic training set is reconstructed according to the penalty function C i and can be controlled dynamically by user, so the training time of the SVM based on target feature can controlled dynamically in a short time; therefore, the new target samples being obtained in the battlefield can applied in the SVM at once. At last, the dynamic recognition method is applied to the underwater target recognition that is utmost important to submarine war. Experiment results show that the dynamic recognition method using SVM is more robust than the traditional SVM.

A RBF (Back-Propogation) neural networks based feature is applied to the target recognition, which aims at only recognition of the target feature and searches the hyperplane of the local space taking the target feature as center. The classifier integrates the target feature with RBF ANNs. It evaluates importance of each sample to the target feature by using expected output of the dynamic ANNs training process. Experiment results show that it is more robust than the traditional method.