超宽带雷达传感器土壤信息获取与处理

Characteristics of soil information include big differences in spatial distribution and large data volume, etc. It is difficult to rely on traditional means of measuring to acquire its spatial and temporal variability in real-time. Theories and technologies in intelligent acquisition and processing of soil information are in urgent need. Ultra-wideband (UWB) radar sensors with high resolution and strong penetrating power are suitable for obtaining information of shallow soil intelligently. This project focuses on: 1) Soil characteristics extraction method via fuzzy logic to explore the inversion mechanism for soil characteristics using UWB echoes. 2) UWB soil channel models via probabilities and statistics to reveal the connection between soil characteristics and UWB channels. 3) Compressive sensing in UWB echoes from soil to solve the bottleneck in high sampling rate resulted from UWB. This project tries to disclose the basic relation between UWB signals and soil characteristics using fuzzy logic and statistical methods, and has a strong scientific significance. This research addresses soil physics, radar detection, and compressive sensing etc., it has a wide cross over several research areas. This project also has broad application prospects. It will promote the development of proprietary soil-information-monitoring and management system, as well as the sustainable development of the precise agricultural, environmental protection, water conservancy and civil engineering, etc.

土壤信息具有空间分布差异大、数据量大等特点，其时空变异性难以依靠传统测量手段实时采集，迫切需要土壤信息获取与处理的智能化理论与技术。超宽带雷达传感器分辨率高、穿透力强, 适易于浅层土壤信息的智能化获取。本项目重点研究：1) 基于模糊逻辑的土壤特征提取方法, 探索超宽带回波的土壤特征反演机理；2) 基于概率统计方法的土壤超宽带信道模型，揭示土壤特征与超宽带信道的作用规律；3) 超宽带土壤回波压缩感知算法，解决带宽太大导致的土壤信息采样速率过高的瓶颈问题。本项目力求通过模糊逻辑与概率统计等方法揭示超宽带信号与土壤特征间的基本联系，具有较强的科学意义；研究内容涉及土壤物理、雷达探测、压缩感知等多个领域，具有较广的交叉性；本项目的实施能促进开发具有自主知识产权的土壤信息监测与管理系统，推动精准农业、环境保护、土木水利等领域的可持续性发展，具有广泛的应用前景。

超宽带雷达传感器因其分辨率高、穿透力强、功耗低等优点，已被用于浅层土壤中的隐蔽目标探测，是一种性能高、成本低、实时性强的技术。但智能化获取土壤含水量、pH值的研究因需要大量实测数据的支持，需要土壤物理学、雷达探测、人工智能等多门学科的交融，因而不易开展。本项目采集了大量不同种类土壤的超宽带雷达回波信号并进行信号处理，首先提出了基于模糊逻辑模板匹配、基于模糊系统特征提取与机器学习模型、基于时频分析图像和深度学习的3种土壤体积含水量分类算法，实现了根据实测超宽带土壤数据反演土壤含水量类别的功能，平均正确识别率81%以上。其次，将土壤特征的分类问题转化成回归问题，提出4种土壤特征预测模型，输入实测土壤数据，输出预测的体积含水量或pH值，综合样本数量与预测误差性能，PCA-GB在4种模型中表现最优。最后，为了大幅降低信号实时采集数据量并提高系统举一反三的学习能力，提出了2种超宽带土壤回波压缩感知算法与1种迁移压缩感知模型，降低了超宽带土壤的采样速率、相比传统算法显著提高了重构信号的精准性。本项目采用的超宽带雷达传感器可装载于无人机对区域性浅层土壤进行组网探测，提出的土壤信息获取方法避免了传统测量方法对土壤及农作物的破坏，有助于实现土壤信息的无损获取，开辟了土壤特性研究的新途径。项目执行以来，共发表SCI二区及以上论文11篇，其中一区Top期刊(IEEE Internet of Things Journal) 2篇。发表国际期刊论文1篇，国际会议论文22篇，获国际会最佳论文奖1项。专利授权4项。培养博士生1人、硕士生6人，均发表中科院SCI大类二区级以上论文。

内容获取失败，请点击重试