Research on Automatic Obstacle Avoidance and Target Tracking System for Drones Based on Deep Learning

This study presents an in-depth analysis and enhancement of the YOLO (You Only Look Once) deep learning algorithm for improving automatic obstacle avoidance and target tracking in drones. Key modifications to the algorithm’s architecture and training process were implemented to optimize its performance for real-time drone applications. Through rigorous experiments, the modified YOLO model demonstrated a significant increase in obstacle avoidance success, with rates rising up to ${9 9 \%}$, and target tracking accuracy improving by an average of $10 \%$. Response times were notably reduced, achieving as low as 110 ms in optimal conditions. The research also incorporated a comprehensive dataset representing diverse environments, which contributed to the robustness and adaptability of the model. Validation processes confirmed the model’s enhanced precision and reliability across various scenarios. These findings indicate a substantial advancement in drone navigation technology, offering promising prospects for applications in surveillance, search and rescue, and agricultural monitoring. The study sets a new benchmark for real-time, efficient, and accurate autonomous drone systems.