A Decision-Level Fusion Method Based on Convolutional Neural Networks for Remote Sensing Scene Classification

Abstract

Scene classification is a key issue in the field of remote sensing image analysis. In recent years, a lot of scene classification methods have emerged due to the convenient acquisition of high spatial resolution remote sensing images. In the existing literature, the feature-level fusion method is widely used to improve the classification performance. In this paper, we propose a decision-level fusion method based on convolutional neural networks (CNNs) for remote sensing scene classification. The proposed method accomplishes the classification task through two steps. In the first step, a CNN is fine-tuned using the training samples, and the soft-max layer of the CNN is used to obtain the top-N possible classes of each test sample. In the second step, multiple linear support vector machines (SVMs) are trained using the features extracted from the fully-connected layer of a pre-trained CNN. The final class of each test sample is determined by the SVMs from the top-N possible classes obtained in the first step. Experiments on the widely used UC Merced data set and AID data set demonstrate the effectiveness of the proposed method.