Town group extraction method and system based on optical remote sensing image

Abstract

The invention provides a town group extraction method and system based on an optical remote sensing image. The method comprises the following steps: firstly, establishing a training sample library according to optical remote sensing image data and global surface coverage classification data; secondly, establishing a deep convolutional neural network model with an attention mechanism; training thedeep convolutional neural network model with the attention mechanism by using the training sample of the training sample library to obtain a town group extraction model; and finally, obtaining an optical remote sensing image of the to-be-detected area, and extracting the town group. According to the invention, a deep convolutional neural network with an attention mechanism is utilized. The remotesensing image terrain classification information is obtained through deep learning semantic segmentation, deeper information such as shapes, textures and semantics in the remote sensing image is fullyutilized, the accuracy of town group extraction based on the optical remote sensing image is improved, manual participation is not needed, and the real-time performance and efficiency of town group extraction are improved.

Description

technical field [0001] The present invention relates to the technical field of urban agglomeration extraction, in particular to a method and system for extracting urban agglomerations based on optical remote sensing images. Background technique [0002] As the scale and number of urban agglomerations gradually increase, it is very important to understand the spatial evolution of urban agglomerations. With the continuous improvement of the spatial resolution of satellite sensors, the major special project of high-resolution earth observation system has provided a large number of high spatial resolution, high temporal resolution and hyperspectral high-resolution remote sensing data. With the successful launch of the subsequent high-resolution satellites, the continuous and stable supply of high-resolution data, and the gradual improvement of data processing services, the application of high-resolution satellite data has achieved remarkable results, filling the gaps in remote ...

AI Technical Summary

Problems solved by technology

[0003] The existing methods for extracting urban agglomerations mainly have the following deficiencies. First, whether it is early artificial interpretation to analyze satellite images to obtain ground object classification results, or the relatively mature support vector machine or neural network to extract remote sensing images The classification of ground objects only analyzes the relatively shallow spectral information of the image, and does not excavate and utilize the deeper information such as shape, texture, and semantics of the image, which affects the accuracy of the extraction of urban agglomerations; secondly, the above Both methods require a lot of repetitive manual work in the classification of ground objects, which is time-consuming and laborious, and the real-time performance is poor.

Images