Image Recognition System for Pico Satellite Earth Surface Analysis (50 – 75 M)

This paper presents the development of a prototype image recognition system for Pico satellite Earth surface analysis, utilizing a Raspberry Pi 5 as the primary computing unit. A Pi Camera captures images of the Earth's surface, which are then processed using the TensorFlow Lite framework for object detection, specifically identifying fire and smoke. A GPS module is integrated to obtain geographic coordinates of the captured images, ensuring accurate location tracking of detected incidents. For data transmission, the system employs the HC-12 wireless communication module to transfer processed image data from the Pico satellite (Raspberry Pi 5) to a ground station. The ground station, consisting of a Raspberry Pi 3, receives the transmitted data via another HC-12 module, where the detected fire or smoke regions and their corresponding bounding coordinates are displayed on a dashboard using the Tkinter graphical user interface (GUI). This prototype offers a cost-effective and efficient approach for real-time fire and smoke detection from a small satellite platform. By integrating machine learning-based object detection with wireless communication, the system enhances remote monitoring capabilities, potentially contributing to wildfire detection and disaster management applications.