Skip to Main Content
The use of remote sensing images from various sensors is supposed to be able to improve classification accuracies. In this paper, a multiple classifiers system is adopted to fully utilize the complementary information among different data sources. A weighting policy may be applied to fuse knowledge acquired by classifiers according to their classification performances. Based on the past researches, there are some kinds of complex relationship among the classifiers' outputs. It is believe that the classification accuracy will be further improved if these relationships could be modeled properly. Therefore, a neural decision maker is proposed to express their relationships and to determine their weights among classifiers' outputs. Another type of the multisource classifier, neural networks approach, is also introduced. The classification performances of utilizing various multisource classifiers, i.e. neural network approach, multiple classifiers systems weighted by y the conventional Bagging and Boosting algorithms and the proposed method, to the application of multisource remote sensing images classification/ data fusion are demonstrated and compared. Experimental results show that both the neural networks approach and multiple classifiers system can dramatically improve the classification accuracy. In addition, the classification performance of the proposed method is better than that of using neural networks approach. Moreover, the proposed method outperforms the multiple classifiers systems weighted by the conventional Bagging and/ or Boosting algorithms.