基于四旋翼视觉平台的水稻LAI及发育期信息获取方法研究

Rice growth information observation (GIO) has always been an important part of rice production. At present, crop GIOs mainly have small-scale manual observation, image based ground station observation, and large-scale remote sensing observation methods. To overcome the shortcomings of the above three kinds of observation methods, it is of great theoretical and practical significance to study the new methods of crop GIO. The aim of this project is to combine the unmanned aerial vehicle (UAV) and the computer vision technology to propose a new method of obtaining rice growth information, such as the key growth stages and leaf area index (LAI), based on the quadrotor vision platform. The research mainly includes: studying the algorithm of false matching points elimination method in panoramic stitching to realize the accurate stitching of rice panorama images; studying the rice color modeling method based on convex hull modeling to realize the accurate segmentation of rice panorama images; incorporating the previous research to achieve the automatic detection of rice key growth stages; and with the manual observation records of LAI, research on the quantity relation between the rice canopy coverage and LAI during different growth stages in order to realize the accurate inversion of rice LAI.

水稻长势信息观测一直是水稻生产中的一个重要内容。目前，作物长势观测主要有小尺度上的人工观测、地面站图像观测，以及大尺度上的遥感观测方式。为克服以上三类观测方式现存的不足，研究作物在长势信息获取上的新方法具有十分重要的理论和现实意义。本项目旨在结合无人机及计算机视觉技术，提出一种新的基于四旋翼视觉平台的水稻长势信息获取方法，实现对水稻关键发育期、叶面积指数LAI信息的自动观测。研究内容主要包括：研究全景拼接中误匹配点剔除算法，实现水稻全景图像的准确拼接；研究基于凸包建模的作物颜色建模方法，实现水稻全景图像的准确分割；衔接前期研究基础，实现水稻关键发育期的自动检测；结合水稻LAI的人工观测记录，探究水稻在不同发育期内冠层覆盖度与LAI之间的定量关系，实现水稻LAI信息的准确反演。

水稻是我国三大粮食作物之一，水稻栽培在我国粮食生产、粮食安全中起着至关重要的作用。水稻长势信息观测技术的研究、推广应用一直受到广泛关注。目前，水稻长势观测主要有以下三类观测方式：小尺度上的人工观测、地面站图像观测，以及大尺度上的遥感观测。本项目在前期已有作物观测数据和研究基础上，期间重点研究了四旋翼无人机作物图像自动采集，作物图像的质量控制方法、作物图像准确分割方法、基于多分类器级联的水稻关键发育的自动检测方法、基于水稻作物冠层覆盖度的水稻叶面积指数LAI、水稻叶片SPAD信息反演方法等。研究中，在水稻抽穗期的图像自动检测方面取得了重要成果并在权威期刊发表；在水稻作物叶面积指数LAI、叶片SPAD信息等水稻冠层长势信息的反演方法上取得了一定科研进展，并进行了应用尝试；另外，创建水稻作物无人机超低空观测图像新数据集一份。