摘要
为了解决目前脉冲激光测距回波估计方法存在的脉冲响应函数不够平滑、展宽不够理想,导致脉冲激光测距精度较低的问题,提出了卷积神经网络的脉冲激光测距回波估计方法。采用学习权值与偏置常量神经元,组建卷积神经网络结构,通过分析五个构成模块主要功能,完成卷积神经网络优化,利用输入样本分量与神经元权重,架构输入输出的约束条件,经过设定网络层数、卷积核数以及激活函数等指标参数,确定神经元个数,根据学习速率初始数值,取得最优邻域初始数值,依据输入的带标签样本实现卷积与降采样处理,获取最优邻域权重,在激光空气传播速度恒定原理与回波波形非对称高斯分布基础上,通过回波脉冲空间水平分布与时域分布,得到最终回波估计结果。实验结果表明,所提方法的回波信号估计结果与实际结果存在的最大偏差值仅为0.2%,脉冲激光测距精度较高。
In order to solve the problem that the pulse response function of the current pulse laser ranging echo estimation method is not smooth enough and the width is not ideal enough,resulting in low accuracy of pulse laser ranging,a pulse laser ranging echo estimation method of the convolutional neural network is proposed.Use learning weights and bias constant neurons to form a convolutional neural network structure.By analysing the main functions of the five constituent modules,complete the convolutional neural network optimisation using the input sample component—neuron weights to construct the input and output constraints.After setting the index parameters such as the number of network layers,convolution kernels and activation functions,the number of neurons is determined,the optimal initial value of the neighbourhood is obtained according to the initial value of the learning rate,and convolution and downsampling are realised according to the input labelled samples After processing,the optimal neighbourhood weights are obtained.Based on the principle of constant laser air propagation speed and the asymmetric Gaussian distribution of the echo waveform,the final echo estimation result is obtained through the horizontal spatial distribution and the time domain distribution of the echo pulse.Experimental results show that the maximum deviation between the estimated results of the proposed method and the actual results is only 0.2%,and the pulse laser ranging accuracy is high.
作者
樊里略
阮清强
陈佳
FAN Lilue;RUAN Qingqiang;CHEN Jia(School of Information Engineering Zunyi Normal University,Zunyi Guizhou 563003,China)
出处
《激光杂志》
CAS
北大核心
2021年第7期32-36,共5页
Laser Journal
基金
贵州省科技合作计划项目(No.黔科合LH字[2015]7014号)
贵州省科技合作计划项目(No.黔科合LH字[2015]7056号)。
关键词
卷积神经网络
脉冲激光测距
回波估计
光强空间分布
convolutional neural network
pulsed laser ranging
echo estimation
spatial distribution of light intensity
