Development of intelligent remote sensing payloads for CubeSat applications
Authors:
Open Access
- 21 March 2025
- conference paper
- Published by SPIE-Intl Soc Optical Eng
- Vol. 13546, 86-1018
Abstract
The paper describes the development of a remote sensing payload for CubeSat applications. The inclusion of artificial intelligence into remote sensing payload has been pursued in the IRIS (Intelligent Remote sensing and Internet Satellite) and Lilium CubeSat programs. The paper discusses the development of a series of intelligent remote sensing payloads for CubeSats to address the limitations on computational resources, power, and communication bandwidth. From 2018 to 2021, under the IRIS program, an optical remote sensing payload that contains a high focal-length telescope, a CMOS detector, and the data operation board was developed. An edge intelligence for landslide detection and cloud segmentation is investigated by using MobileNet V2 network. The design has been implemented and tested by using an airborne platform. The design is further refined in the Lilium program with enhanced resolution and frame rate. Functions including auto-focus, automatic white balance, and gain control are also realized in the payload. To account for low-illumination scenario, a zero DCE (zero reference deep curve estimation) is utilized. Further, a light-weight U-net is implemented in the FPGA to achieve cloud detection and segmentation and a YOLO-SPD is implemented for object detection. This self-developed payload has been prototyped and is under tests. The payload will be integrated into the 6U Lilium-2 CubeSat and is planned to be launched in 2025.Keywords
This publication has 11 references indexed in Scilit:
- OPS-SAT Spacecraft Autonomy with TensorFlow Lite, Unsupervised Learning, and Online Machine Learning2022
- Towards Robust Cloud Detection in Satellite Images Using U-Nets2021
- Cloud detection methodologies: variants and development—a reviewComplex & Intelligent Systems, 2019
- Cloud-Net: An End-To-End Cloud Detection Algorithm for Landsat 8 Imagery2019
- Deep learning based cloud detection for medium and high resolution remote sensing images of different sensorsarXiv, 2019
- U-Net: Convolutional Networks for Biomedical Image Segmentation2015
- Conditional Generative Adversarial NetsarXiv, 2014
- Generative Adversarial NetworksarXiv, 2014
- A Digital Image Stabilization Method Based on the Hilbert–Huang TransformIEEE Transactions on Instrumentation and Measurement, 2012
- A fast algorithm for bidimensional EMDIEEE Signal Processing Letters, 2005