Development of Relevance Propagation Rule-based Systems and Faster Mask R-CNN for IoT-Enabled Surveillance Drones to Enhance Autonomous Navigation and Collision Avoidance

The study focuses on enhancing autonomous navigation and improving collision avoidance by developing a Relevance Propagation Rule-Based on Faster Mask R-CNN (RPR-FMRCNN) leveraging IoT-enabled surveillance drones to enhance smart decision-making. Autonomous drones are increasingly being used for various tasks, but operating them safely and efficiently in complex environments remains a significant challenge, particularly in avoiding collisions with moving obstacles. Existing approaches struggle with real-time decision-making and are often imprecise and unstable in rapidly changing scenarios. The paper proposes a hybrid system for obstacle recognition to address these issues. IoT is used to gather real-time information, ensuring immediate updates. The rule-based approach prioritizes obstacle relevance to dynamically create safer paths, while FMRCNN accurately identifies obstacles, providing boundaries and segmentation masks for each object. IoT-enabled surveillance drones offer seamless connectivity and data exchange, facilitating continuous environmental updates and informed decision-making. The primary goal is to develop a system capable of autonomously determining optimal navigation routes, accurately identifying and categorizing obstacles, and making well-informed decisions to avoid them. By combining deep neural networks with IoT, the model aims to provide real-time processing with improved precision and efficiency. The study's results show significant improvements in obstacle recognition and navigation, with the system performing better in dynamic environments. The outcomes demonstrate a substantial reduction in crashes, enhancing the overall reliability of drones. Compared to existing methods, the proposed model improves collision avoidance efficiency by 15% and navigation accuracy by 12%, signaling promising advancements in autonomous drone systems for various applications.