An underwater multi-class target classification method based on reliability estimation

Abstract

The invention provides an underwater multi-target classification method based on reliability estimation. First, an underwater multi-target data set is constructed and its power set is given, and then the classification result of each two-class SVM classifier is given, and the calculation The conflict factor and the confidence of each two-class SVM classifier are used to obtain the classification accuracy of each underwater target class that needs to be discriminated. Aiming at the problem that the confidence level of each two-class SVM classifier cannot be determined and the results of multiple two-class SVM classifiers cannot be effectively fused, the present invention uses a Gaussian membership function to represent the reliability of each two-class SVM. factor and use the constructed confidence fusion rule to fuse the output results of each two-class SVM with a confidence factor, so that the multi-class underwater targets can be identified on the basis of increasing the credibility of each binary classifier, and the water quality can be improved. Classification accuracy of multi-class targets.

Description

technical field [0001] The invention belongs to the field of information signal processing, and relates to theoretical methods such as underwater signal processing, support vector machine, D-S evidence theory, and classifier fusion. Background technique [0002] Since the 1980s, due to the extremely important application value of underwater target classification and recognition technology, it has become a hot spot in the field of underwater equipment research. Due to the complex and changeable marine environment and the non-stationarity of marine environmental noise, the classification method of underwater multi-class targets is more difficult than the classification and recognition task of underwater two-class targets. [0003] At present, there are many methods to solve multi-classification problems, such as decision tree method, Bayesian method, artificial neural network algorithm and so on. The robustness of the decision tree method is poor, and the effectiveness of the...

AI Technical Summary

Problems solved by technology

Therefore, when using the multi-classification SVM algorithm, there will be two problems: one is that the weight value of each second-class SVM cannot be determined, that is, the confidence level of each second-class SVM classifier cannot be determined; Effective fusion of results from multiple binary SVM classifiers

Images