大豆冠层三维重建及光照时空分布模型研究

High-quality light distribution of soybean canopy has important significance to improve the yield and quality of grain. In order to reveal the low of temporal-spatial distribution to soybean canopy intelligently and efficiently, the soybean canopy is set for research object, the vision system constructed by 3D laser scanner and color camera is used to acquire multi-source data dynamically, and then the parameters of image registration and fusion vector are optimized based on analyzing 3D structure characteristics of soybean canopy, which establish a basis for 3D structure model of soybean canopy with color characteristic. Based on the established 3D model above, changing laws of light radiation characteristics with different solar azimuth, canopy depth and observation angle are researched, the non-linear relationship between color characteristic and light intensity is analyzed and then the mathematical expression mechanism of light distribution changing with time series is proposed, at last, the model for temporal-spatial distribution of light to soybean canopy is established. Meanwhile, the model is verified and optimized by actually measuring light distribution of soybean canopy in reproductive stage. The achievements of this study will bring theoretical basis for determining the temporal-spatial distribution of light to soybean canopy quickly and selecting soybean ecological breeding index effectively as well as the production practice.

大豆冠层优质的光照时空分布，对提升籽粒产量及品质具有重要意义。为了智能高效地揭示大豆冠层光照时空分布特性，本研究以大豆冠层为研究对象，构建彩色摄像机与三维激光扫描仪相结合的视觉系统，动态采集大豆冠层彩色图像和深度信息的多源数据，分析大豆冠层三维结构特征，优化设计图像配准参数及融合向量，建立具有颜色特征的大豆冠层三维结构模型。在此基础上研究大豆冠层内空间光照辐射特性随不同太阳方位角、冠层深度及观测角度的变化规律，应用智能信息处理技术和数据挖掘技术，解析冠层内空间颜色特征与光照强度的非线性映射关系，提出冠层三维空间光照分布随时间序列变化的数学表达机理，建立大豆冠层光照时空分布模型。同时实际测量大豆冠层在生殖生长期的动态光照分布状况，对大豆冠层光照时空分布模型进行验证与优化。项目成果可以科学快速地确定大豆冠层生理生态性状的光照时空分布特性，为大豆生态育种指标的有效选择和生产实践提供相关理论依据。

按照任务书要求，本项目以寒地大豆为研究对象，重点开展了大豆冠层三维重建、颜色特征与光照强度的变化规律和冠层光照时空分布规律解析三个方面的研究工作。以三个不同品种（富豆6号、抗线9号、抗线13号）的大豆冠层为研究对象，构建了以PMD、Kinect、FastScan，彩色摄像机等传感器为基础的多源数据获取系统，动态采集了大豆冠层彩色图像和深度信息等多源数据，分析了不同生殖生长阶段大豆冠层内空间各部分的颜色、形状等特性；应用Harriss算法优化了图像配准参数及融合向量，快速重建了具有颜色信息的大豆冠层三维结构模型；提出了基于颜色恒常性的彩色图像失真校正方法，计算了大豆冠层图像的红光值R、绿光值G、蓝光值B、最佳颜色特征蓝光标准化值B/(R+G+B)、绿光标准化值G/(R+G+B)、色度、饱和度、强度等多维度颜色特征，研究了其冠层内空间位深的变化规律；在重建的三维模型基础上，提出了大豆株高、株宽等多维度表型参数的计算方法；应用计算几何理论，结合“顶视法”的冠层三维空间区域划分方法，通过自主开发的大豆冠层光照强度智能多路测量仪，实时、准确地获得了大豆冠层不同深度剖面的光照强度，解析了冠层三维空间内的光照强度变化规律，研究了不同生殖生长期大豆冠层内空间颜色特征和表型特征与光照强度之间的关系。项目研究成果为旨在获取最佳光照分布的大豆优良品种选育提供了可靠的数据支持。