Thermal sentinel: Low-earth orbit infrared intelligent system for flying civil aircraft safety

The surveillance and detection of civil aircraft over a wide area has long been a technical challenge, with no available datasets and complete detection methods yet. The first global space-based three-channels thermal infrared flying civil aircraft dataset (TIFAD.v1) is established by this paper, covering 17 months, six continents, with 21,004 aircraft and 1252 contrail aircraft, integrating ADS-B civil aviation data. TIFAD.v1 is a fine-grained dataset of small flying targets with aircraft type, calibrated altitude, ground speed, and track. We investigates the radiative feature of high-altitude flying targets and finds that their thermal radiation peaks are primarily distributed in the 10.3–12.5 μm band. Additionally, a significant altitude-dependent difference in transmittance is observed at 11.72 μm, which helps suppress background interference and enhances the reliability of target detection. And an innovative detection method for wide-area flying target is developed, combined YOLOv11n-based deep learning algorithms with in-situ radiative characteristics including the top of atmosphere radiance, temperature contrast, SCR to enhance detection accuracy. This technology provides an effective and robust new approach for all-weather detection of maneuverable flight targets on a global scale, demonstrating the significant potential of intelligent technology in the field of thermal infrared remote sensing applications. Moreover the in situ data allows for quantitative measurement of the radiative feature of flight targets, providing essential data for research on infrared and atmospheric transmittance. • The first large-scale dataset of global flying aircraft observed from a space-based thermal infrared imaging platform. • The innovative combination of simulation and in-orbit data validated the aircraft detection by 8 to 12.5 μm. • Multi-source data fusion for dim target deep learning model without special texture information limit ability verification.