For Unmanned Aerial Vehicles(UAV),the intelligent video analysis is a key technology in intelligent autonomous control,real-time navigation and surveillance.However,poor UAV wireless links would degrade the quality of...For Unmanned Aerial Vehicles(UAV),the intelligent video analysis is a key technology in intelligent autonomous control,real-time navigation and surveillance.However,poor UAV wireless links would degrade the quality of video communication,leading to difficulties in video analysis.To meet the challenges of packet-loss and limited bandwidth in adverse UAV channel environments,this paper proposes a parameter optimization mechanism for UAV intelligent video analysis.In the proposed method,an Optimal Strategy Library(OSL)is designed to optimize the parameters for video encoding and forward error correction.Adapted to the packet-loss rate and bandwidth in practical UAV wireless network,the proposed OSL can facilitate the encoding of video sequences and the recovery of degraded videos with optimal performance.Experimental results demonstrate that the proposed solution can keep intelligent video analysis working efficiently with adverse UAV wireless links,and is capable of maximizing the inference accuracy of Multi-Object Tracking(MOT)algorithms in various scenarios.展开更多
文摘For Unmanned Aerial Vehicles(UAV),the intelligent video analysis is a key technology in intelligent autonomous control,real-time navigation and surveillance.However,poor UAV wireless links would degrade the quality of video communication,leading to difficulties in video analysis.To meet the challenges of packet-loss and limited bandwidth in adverse UAV channel environments,this paper proposes a parameter optimization mechanism for UAV intelligent video analysis.In the proposed method,an Optimal Strategy Library(OSL)is designed to optimize the parameters for video encoding and forward error correction.Adapted to the packet-loss rate and bandwidth in practical UAV wireless network,the proposed OSL can facilitate the encoding of video sequences and the recovery of degraded videos with optimal performance.Experimental results demonstrate that the proposed solution can keep intelligent video analysis working efficiently with adverse UAV wireless links,and is capable of maximizing the inference accuracy of Multi-Object Tracking(MOT)algorithms in various scenarios.
