Abstract:

As for the problems of calculation, data and recognition algorithm complexity of multi-polarized HRRP in target recognition field, a recognition approach based on Bagging-SVM dynamic ensemble is proposed. Firstly, multi-polarized HRRP’s shift invariant feature vectors are extracted. Then Bagging and feature selection methods based on dynamic mutual information are used to train the base classifier. Finally, base classifier diversity is introduced to carry out selective ensemble. The experiments validates that the method performs well in reducing the data dimensions, calculation and has a better recognition correct rate, at the same time, it lets down the HRRP’s aspect sensitivity.