基于神经网络的四旋翼无人机旋翼长度估计算法

Rotor length estimation algorithm of quadrotor UAV based on neural network

针对多旋翼无人机目标识别困难问题,提出一种基于神经网络的四旋翼无人机旋翼叶片长度估计算法。建立四旋翼无人机旋翼回波模型,仿真分析叶片转速和叶片长度等参数对微多普勒和时域、频域特征的影响。利用Gabor变换得到时频特征,通过Radon变换对时频数据预处理,将时频信息、Radon变换信息作为卷积神经网络输入,将时域、频域信息作为BP神经网络的输入,通过子神经网络特征提取结果信息融合估计出旋翼长度。仿真结果表明,该算法能够精准估计无人机旋翼叶片长度。

Aiming at the difficulty of target recognition for multi-rotor UAVs,a rotor blade length estimation algorithm for four-rotor UAVs based on neural network is proposed.A four-rotor UAV rotor echo model was established,and the effects of parameters such as blade speed and blade length on micro-Doppler and time and frequency domain characteristics were analyzed through simulation.Gabor transform is used to obtain time-frequency features.Time-frequency data is preprocessed by Radon transform.Time-frequency information and Radon transform information are input as convolutional neural network.Time-domain and frequency domain information are used as input of BP neural network.The network feature extraction result information fusion estimates the rotor length.Simulation results show that the algorithm can accurately estimate the length of the UAV rotor blades.