Optical remote sensing scene classification method based on deep twin capsule network

Abstract

The invention discloses an optical remote sensing image scene classification method based on a deep twin capsule network. The method comprises the following steps of 1, deleting an average pooling layer of a deep residual network deep layer and a layer behind the average pooling layer; 2, taking the fine-tuned deep residual network as a feature extractor; 3, features of the input images are extracted respectively, and the obtained features are converted into capsule features; 4, introducing the thought of a twin network, and copying a single deep capsule network into a double deep capsule network to form two identical networks with shared feature extractor parameters; 5, calculating the distance between the two features to represent the similarity degree of the image pairs; and 6, carryingout capsule propagation by using a dynamic routing algorithm to finish image classification. Characteristic space information is stored by utilizing capsules, and a twinning network structure is combined, so that the characteristics have higher identifiability on a remote sensing data set. In addition, a regularization term is added to enhance the robustness of the model.

Description

technical field [0001] The invention relates to an optical remote sensing image scene classification method based on a deep twin capsule network, which is a remote sensing image processing technology. Background technique [0002] With the rapid development of satellite radar and UAV aerial photography technology, the collection of remote sensing images has become more convenient and cheaper, giving us more diversified means to observe our world. The study of remote sensing images is of great significance in the field of remote sensing. Remote sensing images can not only reflect the current state, but also obtain dynamic information based on images from different periods. Such as changes in cities, expansion of roads, and destruction of nature. In addition, the full and accurate use of remote sensing images can accurately identify targets in the event of natural disasters, and can quickly provide information about the target environment during military operations. There a...

AI Technical Summary

Problems solved by technology

[0005] (1) The remote sensing data set has very few samples, so we cannot make full use of deep learning methods for deep data mining of remote sensing images;

[0006] (2) Remote sensing images have multiple angles and multiple directions for a scene, and there are very large differences in angles and positions in each category;

If the images in the test set are flipped, tilted or have any other orientation and angle problems, the performance of the convolutional neural network is not ideal;

[0008] (4) In the convolutional neural network classification model, the introduction of the pool layer usually loses spatial information, and can only learn feature invariance, and does not have the ability to learn features.

In the case of remote sensing data sets with rich direction and position features and a small number of images, the feature representation ability of network learning is less robust, and overfitting is easy to occur.

Images