UAV Remote Pilot Fatigue Detection Technology Based On Depth Perception Technology

In recent years,the Chinese civil aviation and unmanned aerial vehicle(UAV)industries have experienced rapid growth,leading to a significant increase in the number of UAVs and a corresponding rise in the workload and fatigue of UAV operators.In contrast to traditional motor vehicle drivers,UAV operators face diverse work scenarios and frequently interact with controllers.As a result,it is critical to detect and mitigate fatigue-related risks.Unfortunately,current fatigue detection methods suffer from poor real-time performance,low accuracy,high environmental requirements,and cumbersome operation,which makes them challenging to implement in practice.To overcome these limitations,this thesis proposes a fatigue detection method for UAV operators based on depth perception technology.This technology enables real-time monitoring of visual characteristics,such as eye movements,mouth posture,and head position,to accurately and swiftly detect signs of fatigue.When fatigue is identified,an alarm is triggered to prevent potential accidents and ensure the safety of the UAV.This study provides a comprehensive investigation and analysis of the fatigue problem among UAV operators,emphasizing the importance of fatigue detection to ensure operator and public safety.To achieve this,a large-scale UAV operator fatigue detection dataset(OFDD)was constructed and analyzed,revealing fatigue-related characteristics such as eye aspect ratio,mouth opening and closing,and head posture.Building on these findings,our proposed method enhances the YOLOv5 network,balances multi-scene data,incorporates the PFLD module to extract eye and mouth features,and models head posture using depth perception technology to extract head posture features.The method also adjusts the network input and anchor frame size,introduces timing features,and employs the Kinect camera as the image acquisition device.This device is unaffected by working conditions and lighting and enables fatigue detection in both day and night conditions.The experimental results demonstrate that the proposed method achieves high accuracy and fast response times in detecting fatigue among UAV operators,thus highlighting its effectiveness in mitigating fatigue-related risks.This study also suggests optimization schemes,such as building large datasets and optimizing the algorithm structure,to further improve the detection accuracy and response speed.These findings have important theoretical and practical implications for enhancing the safety and stability of UAVs.