Remote sensing image scene recognition method and device

Abstract

The invention provides a remote sensing image scene recognition method and device, and belongs to the technical field of image recognition. The method comprises the following steps: depth features ofa remote sensing image are extracted based on a pre-trained deep convolutional neural network; SIFT features of the remote sensing image are extracted; a scene type of the remote sensing image is determined according to the SIFT features and the depth features. According to the remote sensing image scene recognition method disclosed in the present invention, the depth features of the remote sensing image are extracted based on the pre-trained deep convolutional neural network. The SIFT features of the remote sensing image are extracted. The scene type of the remote sensing image is determinedaccording to the SIFT features and the depth features. Because SIFT features have scale invariance and rotation invariance, a problem that the depth features are sensitive to remote image rotation transformation or scale transformation can be overcome during processes of identifying the scene type of the remote sensing image, and therefore identification accuracy of the scene type of the remote sensing image can be improved.

Description

technical field [0001] The present invention relates to the technical field of image recognition, and more specifically, to a remote sensing image scene recognition 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. Based on the earth's surface data obtained by Gaofen satellite, disaster monitoring, agricultural production estimation and military investigation can be realized. Usually, the data frames transmitted from high-resolution satellites to the ground are huge. In order to make full use of these large amounts of high-resolution remote sensing data, scene classification is a very important preprocessing method. Scene classification of these high-resolution remote sensing data can be of great help to subsequent tasks such as target recognition, change detection, and image retrieval. The scene recognition metho...

AI Technical Summary

Problems solved by technology

[0006] Since remote sensing images usually have the characteristics of changeable rotation and scale, when performing scene recognition based on features, if the remote sensing image is rotated, scaled or changed in brightness, it will affect the recognition result, resulting in the difficulty of recognizing the remote sensing image scene. is less accurate

Images