Аңдатпа
Small unmanned aerial vehicles (UAVs) are increasingly used for wildfire detection, where they must not only identify fire events rapidly but also transmit large volumes of sensor data securely to ground stations. Achieving both fast on-board analysis and high-speed encrypted data transmission within the size, weight, and power limits of UAV platforms remain a major technical challenge. In this study, we introduce a compact, FPGA-based system that simultaneously performs real-time fire detection and high-throughput data encryption. Our system integrates a programmable logic chip (FPGA), deep-learning models for visual recognition, and AES-256 cryptographic cores onto a single hardware module. A key innovation is a shared scheduling mechanism that coordinates these two functions efficiently. Furthermore, we demonstrate how artificial intelligence contributes beyond image classification: a lightweight neural network monitors input data streams and dynamically adjusts encryption key parameters, thereby improving security without compromising performance. The hardware supports encrypted data transfer rates of 800 megabits per second at a latency of just 2 microseconds, while identifying fire signatures at 30 frames per second. Extensive testing, including cross-validation on a 50,000-frame dataset and environmental stress testing from –20 °C to 55 °C, confirms robust performance under real-world conditions. While the current memory footprint limits multi-camera input, this work offers a foundational design for future systems that aim to combine edge computing, secure communications, and AI-driven perception in autonomous aerial platforms.