Remote sensing image labeling system and method

Abstract

The embodiment of the invention discloses a remote sensing image annotation system and method. The system comprises a user management module which manages user authority, confirms user login information and determines a user role; a data interaction module which uploads the remote sensing image to a remote sensing image labeling system according to administrator data uploading operation; a slicing module which slices the remote sensing image to obtain a tile image data set; an annotation task publishing module which splits the tile image data set into a plurality of to-be-annotated tasks according to an administrator task publishing instruction and distributes the to-be-annotated tasks to annotators; an image annotation module which annotates the tile image of the to-be-annotated tasks and sends the tile image to an auditor for auditing; a data storage module which is used for storing the annotated images passing the auditing; and a data interaction module which downloads the stored marked image according to the data downloading operation of the administrator. According to the scheme of the embodiment, high-efficiency annotation on a large batch of wide-breadth remote sensing images is realized, and high-quality image annotation with a unified standard is realized.

Description

Technical field [0001] This article relates to graphical processing techniques, especially in remote sensing images labeling systems and methods. Background technique [0002] In recent years, deep learning is greatly applied to the visual tasks such as target testing, semantic segmentation, and has great application value for remote sensing image intelligent analysis. To get a valid depth learning model, you are inseparable from high quality labels. However, creating remote sensing images with tags is a cost high business. At present, the image is constructed from the remote sensing image, and the image is taken from the GIS (Geographic Information System) platform such as Google Earth, and then uses common labeling tools such as Labelme, ArcGIS, which are annotated, this labeling method, There is at least the following questions: [0003] 1. Number of labeling is not uniform, lacking the quality audit process of labeling results; [0004] 2, the size is small, through a single ...

AI Technical Summary

Problems solved by technology

[0003] 1. The labeling specifications are not uniform, and there is a lack of quality review process for labeling results;

[0004] 2. The processing scale is small, and it cannot process large-scale and wide-format remote sensing images through stand-alone tool annotation;

[0005] 3. The marking system has a single tool, which can only be clicked by the marking personnel operating the annotator, lacking effective computer-aided tools

Images