In order to achieve a rapid and accurate identification of soil stratification information and accelerate the development of smart agriculture,this paper conducted soil stratification experiments on agricultural soils...In order to achieve a rapid and accurate identification of soil stratification information and accelerate the development of smart agriculture,this paper conducted soil stratification experiments on agricultural soils in the Mollisols area of Northeast China using Ground Penetrating Radar(GPR)and obtained different types of soil with frequencies of 500 MHz,250 MHz,and 100 MHz antennas.The soil profile data were obtained for 500 MHz,250 MHz,and 100 MHz antennas,and the dielectric properties of each type of soil were analyzed.In the image processing procedure,wavelet analysis was first used to decompose the pre-processed radar signal and reconstruct the high-frequency information to obtain the reconstructed signal containing the stratification information.Secondly,the reconstructed signal is taken as an envelope to enhance the stratification information.The Hilbert transform is applied to the envelope signal to find the time-domain variation of the instantaneous frequency and determine the time-domain location of the stratification.Finally,the dielectric constant of each soil horizon is used to obtain the propagation velocity of the electromagnetic wave at the corresponding position to obtain the stratification position of each soil horizon.The research results show that the 500 MHz radar antenna can accurately delineate Ap/Ah,horizon and the absolute accuracy of the stratification is within 5 cm.The effect on the soil stratification below the tillage horizon is not apparent,and the absolute accuracy of the 250 MHz and 100 MHz radar antennas on the stratification is within 9 cm.The overwhelming majority of the overall calculation errors are kept to within 15%.Based on the three central frequency antennas,the soil horizon detection rate reaches 93.3%,which can achieve accurate stratification of soil profiles within 1 m.The experimental and image processing methods used are practical and feasible;however,the GPR will show a missed detection for soil horizons with only slight differences in dielectric properties.Overall,this study can quickly and accurately determine the information of each soil stratification,ultimately providing technical support for acquiring soil configuration information and developing smart agriculture.展开更多
基于视频的目标检测在恶劣天气情况下识别效果较差,故需弥补视频缺陷、提高检测框架的鲁棒性。针对此问题,文中设计了一个基于雷达和视频融合的目标检测框架,利用YOLOv5(You Only Look Once version 5)网络获得图片特征图与图片检测框,...基于视频的目标检测在恶劣天气情况下识别效果较差,故需弥补视频缺陷、提高检测框架的鲁棒性。针对此问题,文中设计了一个基于雷达和视频融合的目标检测框架,利用YOLOv5(You Only Look Once version 5)网络获得图片特征图与图片检测框,利用基于密度的聚类获得雷达检测框,并将雷达数据进行编码,得到基于雷达信息的目标检测结果。最后将两者的检测框叠加得到新ROI(Region of Interest),并得到融合雷达信息后的分类向量,提高了在极端天气下检测的准确率。实验结果表明,该框架的mAP(mean Average Precision)达到了60.07%,且参数量仅为7.64×10^(6),表明该框架具有轻量级、计算速度快、鲁棒性高等特点,可以被广泛应用于嵌入式与移动端平台。展开更多
针对已有Cohen类时频分布等方法时频聚焦能力不足、在低信噪比(signal to noise ratio,SNR)情况下调制识别准确率低的问题,提出一种基于同步提取变换(synchro-extracting transform,SET)去噪的分组卷积神经网络调制识别方法。所提方法使...针对已有Cohen类时频分布等方法时频聚焦能力不足、在低信噪比(signal to noise ratio,SNR)情况下调制识别准确率低的问题,提出一种基于同步提取变换(synchro-extracting transform,SET)去噪的分组卷积神经网络调制识别方法。所提方法使用SET对雷达信号进行时频分析,以获得良好的时频聚焦性,提高时频分析的计算效率;通过Viterbi算法搜索估计时频系数矩阵中的瞬时频率轨迹,综合考虑信号能量强度分布与瞬时频率轨迹的平滑性,并对得到的瞬时频率轨迹进行中值滤波以去除脉冲噪声;保留瞬时频率轨迹邻域的时频系数,以达到时频图去噪的目的。最后,将去噪后的时频图送入具有残差连接的分组卷积神经网络进行特征提取与调制识别。实验结果表明,当SNR为-12 dB时,去噪后的SET时频图时频聚焦性好,调制识别准确率比未去噪的识别准确率提高了13.69%,证明所提出的雷达信号调制识别方法在低SNR条件下对多种复杂调制类型的信号具有良好的识别性能。展开更多
雷达波形设计逐渐成为现代雷达领域研究的热点,本文首先提出一种OFDM(Orthogonal Frequency Division Multiplexing,正交频分复用)混沌随机相位编码信号,并在此基础上用混沌序列对子载波进行频率调制,得到一种OFDM随机相位随机频率编码...雷达波形设计逐渐成为现代雷达领域研究的热点,本文首先提出一种OFDM(Orthogonal Frequency Division Multiplexing,正交频分复用)混沌随机相位编码信号,并在此基础上用混沌序列对子载波进行频率调制,得到一种OFDM随机相位随机频率编码雷达信号形式,分析该信号的模糊函数、自相关特性以及子载波数对自相关性能的影响。同时,对该混沌编码信号存在的问题进行分析,并研究降低PMEPR(Peak to Mean Envelope Power Ratio,包络峰均比)的方法。仿真结果表明:基于混沌序列的OFDM相位频率组合调制信号具有更优的模糊函数和自相关旁瓣,混沌序列的复杂性和随机性使得信号具有更强的低截获特性。展开更多
基金Under the auspices of the National Key R&D Program of China(No.2021YFD1500100)the Strategic Priority Research Program of the Chinese Academy of Sciences(No.XDA28100000)。
文摘In order to achieve a rapid and accurate identification of soil stratification information and accelerate the development of smart agriculture,this paper conducted soil stratification experiments on agricultural soils in the Mollisols area of Northeast China using Ground Penetrating Radar(GPR)and obtained different types of soil with frequencies of 500 MHz,250 MHz,and 100 MHz antennas.The soil profile data were obtained for 500 MHz,250 MHz,and 100 MHz antennas,and the dielectric properties of each type of soil were analyzed.In the image processing procedure,wavelet analysis was first used to decompose the pre-processed radar signal and reconstruct the high-frequency information to obtain the reconstructed signal containing the stratification information.Secondly,the reconstructed signal is taken as an envelope to enhance the stratification information.The Hilbert transform is applied to the envelope signal to find the time-domain variation of the instantaneous frequency and determine the time-domain location of the stratification.Finally,the dielectric constant of each soil horizon is used to obtain the propagation velocity of the electromagnetic wave at the corresponding position to obtain the stratification position of each soil horizon.The research results show that the 500 MHz radar antenna can accurately delineate Ap/Ah,horizon and the absolute accuracy of the stratification is within 5 cm.The effect on the soil stratification below the tillage horizon is not apparent,and the absolute accuracy of the 250 MHz and 100 MHz radar antennas on the stratification is within 9 cm.The overwhelming majority of the overall calculation errors are kept to within 15%.Based on the three central frequency antennas,the soil horizon detection rate reaches 93.3%,which can achieve accurate stratification of soil profiles within 1 m.The experimental and image processing methods used are practical and feasible;however,the GPR will show a missed detection for soil horizons with only slight differences in dielectric properties.Overall,this study can quickly and accurately determine the information of each soil stratification,ultimately providing technical support for acquiring soil configuration information and developing smart agriculture.
文摘基于视频的目标检测在恶劣天气情况下识别效果较差,故需弥补视频缺陷、提高检测框架的鲁棒性。针对此问题,文中设计了一个基于雷达和视频融合的目标检测框架,利用YOLOv5(You Only Look Once version 5)网络获得图片特征图与图片检测框,利用基于密度的聚类获得雷达检测框,并将雷达数据进行编码,得到基于雷达信息的目标检测结果。最后将两者的检测框叠加得到新ROI(Region of Interest),并得到融合雷达信息后的分类向量,提高了在极端天气下检测的准确率。实验结果表明,该框架的mAP(mean Average Precision)达到了60.07%,且参数量仅为7.64×10^(6),表明该框架具有轻量级、计算速度快、鲁棒性高等特点,可以被广泛应用于嵌入式与移动端平台。
文摘雷达波形设计逐渐成为现代雷达领域研究的热点,本文首先提出一种OFDM(Orthogonal Frequency Division Multiplexing,正交频分复用)混沌随机相位编码信号,并在此基础上用混沌序列对子载波进行频率调制,得到一种OFDM随机相位随机频率编码雷达信号形式,分析该信号的模糊函数、自相关特性以及子载波数对自相关性能的影响。同时,对该混沌编码信号存在的问题进行分析,并研究降低PMEPR(Peak to Mean Envelope Power Ratio,包络峰均比)的方法。仿真结果表明:基于混沌序列的OFDM相位频率组合调制信号具有更优的模糊函数和自相关旁瓣,混沌序列的复杂性和随机性使得信号具有更强的低截获特性。
