Multi-height Visual Drone Positioning Based on LSTM and Convolutional Neural Networks

The ability to autonomously and precisely locate unmanned aerial vehicles (UAVs) is critical to successfully operate in complex and challenging environments. This paper addresses the challenge of location determination for UAVs in scenarios where GPS signals are weak or unavailable. The proposed solution introduces a novel multi-height localization system, leveraging the power of Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNNs) to process visual data captured by a UAV's onboard camera. By analyzing visual information, this system enables UAVs to determine their positions at various altitudes accurately. When GPS signals are unreliable or obstructed, the proposed method offers a robust alternative, enhancing the overall reliability and autonomy of UAV missions. Experimental results demonstrate the real-time effectiveness of our multi-height localization system, showcasing its capability to accurately determine UAV locations at different altitudes.