多元信息融合的农田土壤孔隙度快速评价方法研究

Soil porosity is one of the main physical characteristics of soil. Therefore, it is necessary to find a rapid and efficient evaluation method for soil porosity. Based on the analysis of the interrelation and restriction factors of various soil physical indexes, the information of soil conductivity obtained in real time and soil surface roughness obtained based on image processing technology was fused to get rapid evaluation method of soil porosity. The main contents of this paper are as follows: the coupling relationship between soil physical indexes based on soil conductivity inversion is analyzed. Based on 3D image representing soil surface roughness and then retrieving soil surface porosity, Kinect camera is used to obtain scene depth information quickly and accurately, and 3D point cloud world coordinates are extracted, and roughness information is mined and interpreted. To find a relatively reliable and accurate rough equivalent index variable mapping soil porosity variation, and analyze the relationship between soil conductivity 3D image and soil surface roughness / soil surface porosity. A soil porosity prediction model based on multivariate information fusion was established. The results will provide theoretical support for the development of soil porosity parameters and rapid acquisition system, and provide the basis for fine management of farmland.

土壤孔隙度是土壤的主要物理特征之一，由于现有测量方法无法满足实时信息获取需求，因而寻找快速高效土壤孔隙度评价方法非常必要。本项目在分析各种土壤物理指标相互联系和制约因素基础上，确定从实时获取的土壤电导率以及基于图像处理技术获取土壤表面粗糙度的信息融合入手，研究农田土壤孔隙度快速评价方法。主要研究内容：基于土壤电导率反演土壤孔隙度状况时土壤各物理指标间耦合关系影响分析；基于3D图像表征土壤表面粗糙度进而反演土壤表层孔隙度，采用Kinect相机快速准确地获取场景深度信息并进行3D点云世界坐标提取，对粗糙度信息进行数据挖掘与解译基础上，寻找相对可靠、准确的粗糙当量指数变量映射土壤孔隙度差异变化；分析土壤电导率，3D图像与土壤表面粗糙度/土壤表层孔隙度之间关系，建立基于多元信息融合的土壤孔隙度预测模型。研究结果将为开发土壤孔隙度参数无损、快速获取系统提供理论支撑，为农田精细管理提供依据。

本项目进行了多元信息融合的农田土壤孔隙度快速评价方法研究，结果表明，基于土壤电导率反演土壤孔隙度状况时，土壤水分，盐分，质地等都产生了一定的影响，需要综合考虑这些因素对土壤孔隙度状况反演的影响。同时进行了可见光图像与深度信息融合算法的探究，土壤表面的纹理特征和深度信息融合，可以获得土壤表面的凹凸状况信息，进而对土壤表面粗糙度进行定量或定性的描述。基于传统的最近点迭代（Iterative closest point，ICP）算法对配准点云的空间位置要求苛刻的问题，提出了改进的最近点迭代算法，对得到的土壤表面图像进行处理，提高了配准的精度，从而更好实现了土壤表面三维重建。通过可见光图像、深度图像的配准和融合结合土壤电导率及土壤紧实度等理化参数，构建了土壤孔隙度预测模型。解决了田间复杂环境干扰下，土壤孔隙度自动提取的问题。为了实现大田环境下土壤孔隙度的检测，开发了基于树莓派的车载式土壤容重及孔隙度检测系统，大大降低了人工试验的成本。更进一步，为了实现土壤容重及孔隙度的自动检测，开发了基于云服务的表土层土壤多参数分析与远程管理系统软件，旨在对土壤表层容重、孔隙度等信息进行大田环境下实时获取，并将结果同步上传至软件平台。该项研究为土壤多参数自动检测提供了技术支持。