摘要
与载人飞机相比,商用无人机具有体积小、造价低、使用方便、对作战环境要求低、战场生存能力强等优点,备受多方面的青睐。面对复杂的飞行环境,遥控器、地面站等传统控制系统已无法满足应用要求。考虑到用户使用习惯和无人机控制的可及性,本文提出了一种基于数据手套的无人机端侧控制系统和一种实时的动态手势分割算法。首先通过设置角速度阈值,分割长短不一的动态手势;然后采用线性插值法和组合数据均值法,将长短不一的动态手势重新采样至定长;最后还提出了一种多流一维卷积与Transformer相结合的手势识别算法,在测试集与户外实验上分别取得了98.76%和97.78%的高识别率,这比传统LSTM算法的识别率分别提高了3.39%和5.56%,表明该算法不仅具有良好的识别性和泛化能力,且具有良好的实时性和快速响应能力,户外实验中单次手势完成后0.8s内,无人机便可以执行相应指令,展现出了良好的应用潜力。
Compared with manned aircraft,unmanned aerial vehicles(UAVs)are favored by militaries around the world because of their small size,low cost,convenient use,low requirements on combat environment and strong battlefield survivability.In the face of complex flight environment,traditional control systems such as remote control and ground station can’t meet the application requirements.Considering the user's habit and the accessibility of UAV control,this paper proposes an endto-end control system of UAV based on data glove and a real-time dynamic gesture segmentation algorithm.Firstly,dynamic gestures of different lengths were divided by setting the threshold of angular velocity.Then using linear interpolation method and combined data mean method,dynamic gestures of different lengths were resampling to fixed length.Finally,a gesture recognition algorithm combined with multi-flow one-dimensional convolution and Transformer is proposed,which achieves a high recognition rate of 98.76%and 97.78%in the test set and outdoor experiment,respectively,which is 3.39%and 5.56%higher than the traditional LSTM algorithm.It shows that the algorithm not only has good recognition and generalization ability,but also has good real-time and rapid response ability.In the outdoor experiment,the UAV can execute the corresponding command within 0.8s after the completion of a single gesture,showing good application potential.
作者
陶烨豪
方建波
尚杰
TAO Yehao;FANG Jianbo;SHANG Jie(Faculty of Electrical Engineering and Computer Science,Ningbo University,Ningbo 315211,China;Ningbo Institute of Materials Technology&Engineering,Chinese Academy of Sciences,Ningbo 315201,China)
出处
《无线通信技术》
2024年第1期42-46,共5页
Wireless Communication Technology
