High-resolution thermal infrared contrails images identification and classification method based on SDGSAT-1

Abstract

With the rapid development of the aviation industry, the impact of contrails on global warming is becoming increasingly significant. However, due to limitations in sensor technology and data sources, precise, all-weather observations of contrails remain a significant challenge. To address this, this study utilizes thermal imagery data from the SDGSAT-1 to classify newly formed contrails based on their length, creating SDGCONTR, the first three-band thermal infrared contrail dataset with a 30-meter spatial resolution. Based on this dataset, we propose an effective contrail image classification method: the SDGContrail Images Classification Method. This method builds on the existing Multi-axis Vision Transformer deep learning model. In the preprocessing stage, it employs an innovative contrail enhancement method by creating masks based on contrail brightness temperature and brightness temperature ratio percentiles in thermal infrared images. Our method achieved optimal classification performance compared to other methods, with the F1 Score of 95.2%. Using this model, we further analyzed the geographic and diurnal distribution patterns of long and short contrails, validating the mechanism of contrail formation. Our method not only enhances the ability to conduct large-scale, all-weather, and detailed contrail observations, but also sets a new standard for environmental monitoring and aviation safety.