[P-CG20] Future missions and instrumentation for space and planetary science

[PCG20-14] A 1U CubeSat System Integrating Embedded AI and a Miniature Hyperspectral Spectrometer for Real-Time Image Classification and Cloud Coverage Detection

*Bing-Chen Lai^1,3, Wen-Qian Chang^1,2, Pei-Yuan Li¹, Ming-Ta Hsieh^1,2, Cheng-Ling Kuo^1,2 (1.Center for Astronautical Physics and Engineering, National Central University, Taoyuan City 320317, Taiwan, 2.Department of Space Science and Engineering, National Central University, Taoyuan City 320317, Taiwan, 3.Interdisciplinary Program of Earth System Sciences, National Central University, Taoyuan City 320317, Taiwan)

Keywords:CubeSat, Embedded AI, Hyperspectral Imaging, Real-Time Classification, Cloud Coverage Detection, Low-Power Computation

With the rapid advancement of CubeSat and AI technologies, the integration of real-time AI image recognition with spectral analysis has opened up new application opportunities for small satellites. This study proposes a 1U CubeSat system that combines an embedded AI module with a miniature hyperspectral spectrometer, enabling on-orbit real-time image classification, anomaly detection, and cloud cover labeling. This approach effectively enhances the value of satellite imagery and reduces data transmission requirements.
By comparing onboard observations with real ground data, the system quickly identifies anomalous regions and downloads corresponding RGB and hyperspectral images to support timely assessment of environmental changes and the development of response strategies on the ground. The system provides five primary classification labels—clouds, water, ice, buildings, and land. In particular, cloud coverage can be analyzed in real time at different levels of obscuration, and images deemed overly cloud-covered are automatically excluded to avoid transmitting unnecessary data. Achieving efficient image classification under low power consumption and effectively overcoming AI computing constraints in a 1U CubeSat platform.