Remote sensing image target extraction system and method based on deep learning

Abstract

The invention discloses a remote sensing image target extraction system and method based on deep learning, and the system comprises: a backbone network module which is used for carrying out the downsampling of an original image for many times, and obtaining a first low-level feature, a second low-level feature, a third low-level feature and a fourth low-level feature; the discrimination context perception feature extraction module that is used for obtaining a multi-scale context feature difference fusion result according to the fourth low-level feature; the first up-sampling module that is used for obtaining a first advanced feature according to a multi-scale context feature difference fusion result; the first refining decoder module that is used for fusing and up-sampling the third low-level feature and the first high-level feature to obtain a second high-level feature; and the second refining decoder module that is used for fusing and up-sampling the second low-level feature and thesecond high-level feature result to obtain a third high-level feature. The remote sensing image target extraction system and the remote sensing image target extraction method can enhance the identification capability of background and target features, and have good target extraction capability.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a remote sensing image target extraction system and method based on deep learning. Background technique [0002] Automatic extraction of artificial targets is one of the main tasks of remote sensing systems, and it has important practical significance in applications such as urban planning, geographic information system upgrades, intelligent transportation systems, disaster monitoring, emergency response, illegal building surveys, and geographic information systems. Due to the characteristics of remote sensing scenes such as cluttered background, large differences in target appearance, and radiation distortion, remote sensing image target extraction is a very challenging task. [0003] Object extraction can be viewed as a binary pixel-level classification task, where objects such as buildings, roads, or vehicles are segmented from the surrounding background. ...

AI Technical Summary

Problems solved by technology

Therefore, these simple clustering strategies may degrade the prediction of object details, especially for small and inconspicuous objects

Finally, most of the previous methods for object extraction from remote sensing images have been developed and validated only for specific classes of objects (such as buildings or roads), and their performance may degrade significantly on other object datasets of different classes

Images