Remote sensing image interpretation method based on adaptive sample set construction and deep learning

Abstract

The invention discloses a remote sensing image interpretation method based on adaptive sample set construction and deep learning. The method comprises the steps: carrying out the feature extraction of a total sample set, carrying out the clustering of the extracted features, constructing a feature dictionary of a visual word bag model, and obtaining a sample subset; constructing an interpretation model based on a deep learning network, and successively inputting the total sample set and the clustered sample subsets to train the interpretation model to respectively obtain a total interpretation model and sub-interpretation models corresponding to the sample subsets; and adopting the total interpretation model and a plurality of appropriate sub-interpretation models selected according to the image features of the to-be-interpreted remote sensing image to perform adaptive interpretation on the to-be-interpreted remote sensing image. The method has the remarkable effects that the sample library of the massive remote sensing images is quickly established through automatic and distributed clustering means, the data of the sample library is trained by utilizing a machine deep learning technology, the intelligent interpretation model suitable for different scenes is obtained, the interpretation precision is high, and the robustness is good.

Description

technical field [0001] The invention relates to the technical field of remote sensing image processing, in particular to a remote sensing image interpretation method based on adaptive sample set construction and deep learning. Background technique [0002] After decades of development in remote sensing technology, the resolution of satellite remote sensing images has been continuously improved. With the successful launch of the domestic resources No. 3, Gaofen series and other commercial satellites, remote sensing information extraction has obtained sufficient information sources. The classification of remote sensing images is an important part of obtaining land cover information by means of remote sensing technology. How to use the big data of remote sensing images to realize automatic classification of remote sensing images and extraction of image change information in different time phases is a very important research topic. [0003] In recent years, the use of high-reso...

AI Technical Summary

Problems solved by technology

[0005] However, the interpretation of the above two methods has relatively large limitations. Due to the lack of universal principles or theories in image segmentation and feature selection, traditional pattern recognition methods are difficult to achieve the requirements required for practical applications in large-scale complex images. Accuracy and robustness; in the high-reliability information extraction of remote sensing images by artificial intelligence methods, the existing computer vision methods represented by deep learning are also difficult to match the robustness of the human visual system, mainly due to the lack of a large number of labeled The pixel-level professional database is available for analysis and research, so how to effectively construct and utilize the existing database is imminent

Images