A method and a system for classifying all-sky aurora images and locating key local structures

Abstract

The invention belongs to the technical field of data identification and data representation, and discloses a method and a system for classifying all-sky aurora images and locating key local structures. In the training stage, an image block scale model is constructed on all-sky aurora image data with image labels. The image label is transformed into the region boundary box label by using the imageblock scale model. On the all-sky aurora image database with boundary box label, the image region scale model is trained. In the reasoning stage, the input aurora images are classified by using the image region scale model and the boundary box level rough location of the key local structure is performed. The pixel-level localization of the key local structure of all-sky aurora images is further carried out by using the image block scale model. The invention uses all-sky aurora images with image-level labels as training samples, obtains good image-level classification and pixel-level key localstructure positioning effect, and can be used for automatic analysis of aurora morphology and spatial position evolution process in all-sky aurora images.

Description

Technical field [0001] The invention belongs to the technical field of data recognition and data representation, and in particular relates to a method and system for classification of all-sky aurora images and key local structure positioning. Background technique [0002] Aurora is produced by the excitation of molecules or atoms in the upper atmosphere by high-energy charged particles from the geomagnetosphere or the sun. It reflects changes in solar activity and the Earth’s magnetosphere, and is an important means of monitoring and exploring physical processes in near-Earth space. Among many auroral observation equipment, the ground-based optical all-sky imager can capture the two-dimensional shape information of aurora, and has good spatial and temporal resolution. The all-sky auroral images taken by it are widely used to study the physical evolution of aurora. Since the shape of aurora is related to specific magnetospheric state and dynamic activity, and is affected by solar ...

AI Technical Summary

Problems solved by technology

[0004] In summary, the problems in the prior art are: Existing technologies such as classification and segmentation of all-sky aurora images cannot simultaneously realize the classification of aurora images and the location of key local structures; most weakly supervised semantic segmentation methods cannot be directly used for automatic analysis of all-sky aurora images; they can be directly applied to The weakly supervised semantic segmentation method for all-sky auroral images cannot capture the overall morphological features and local detail differences at the same time, resulting in low classification and key local structure positioning accuracy

Images