Aircraft recognition in remote sensing images based on deep learning

Object recognition is one of the fundamental issues in the field of computer vision. In traditional methods, invariant features are extracted from segmented targets for recognition. However, there is no common method for segmentation of aircraft targets so far due to the complex backgrounds, illuminations, noise and other practical factors. Therefore, in this paper, we propose a method for aircraft identification in remote sensing images based on HOG and deep learning features. We train two classifiers, one is the SVM classifier based on HOG feature, and the other is a classifier based on deep convolutional neural network VGGNet. First, we use the SVM classifier to identify the aircraft in the picture roughly, then we use the deep learning classifier to exclude misidentified targets. In this way, this coarse to fine framework can significantly improve the speed and accuracy of aircraft recognition in remote sensing images. At the same time, our method has a better generalization capability than the traditional methods. Experimental results demonstrate the robustness of our method.