A remote sensing image scene classification method and device

Abstract

The present invention provides a remote sensing image scene classification method and device, the method comprising: based on the compressed convolutional neural network, extracting the picture feature vector of the input remote sensing image picture; based on the picture feature vector, using clustering support vector machine to The pictures are classified into scenes. A remote sensing image scene classification method and device proposed by the present invention reduces the model parameters of the convolutional neural network by compressing the convolutional neural network, and significantly reduces the storage space occupied without reducing the accuracy of scene classification. And the strategy of clustering first and then classification is adopted, which speeds up the classification speed and improves the classification efficiency.

Description

technical field [0001] The present invention relates to the technical field of information processing, and more specifically, to a remote sensing image scene classification method and device. Background technique [0002] With the development of my country's aerospace technology, more and more high-resolution satellites are launched into space to obtain data on the earth's surface. The high-resolution satellites launched into space can play a role in disaster monitoring, agricultural production estimation, and military investigation. Usually, the data transmitted from high-resolution satellites to the ground has a huge frame size. In order to make full use of these large and large high-resolution remote sensing data, scene classification is a very important preprocessing method. It is necessary for further target recognition, change detection, and image retrieval. It is of great help for other tasks. [0003] At present, there are mainly the following methods for remote sen...

AI Technical Summary

Problems solved by technology

However, for remote sensing images, the classification accuracy is still not up to the expected goal, because most of the models currently used for this task are trained on natural image datasets, and the pertinence for remote sensing images is not strong. And because there are few feature parameters in the data set of remote sensing images, while the convolutional neural network has many parameters, large models, and high computational consumption, it is easy to cause overfitting of remote sensing images, so it is difficult to store and store them on mobile devices. well applied to practical tasks

Images