水肥胁迫对基于时间序列图像的苗木表型信息的影响  

Effects of Water and Fertilizer Stress on Seedling Phenotypic Information Based on Time Series Images

作　　者：杨琨琪 张慧春[1] 张萌 边黎明 Yang Kunqi;Zhang Huichun;Zhang Meng;Bian Liming(College of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037;Co-Innovation Center for Sustainable Forestry in Southern China Key Laboratory of Forest Genetics&Biotechnology of Ministry of Education College of Forestry Nanjing Forestry University,Nanjing 210037)

机构地区：[1]南京林业大学机械电子工程学院,南京210037 [2]南京林业大学林学院、南方现代林业协同创新中心、林木遗传与生物技术教育部重点实验室,南京210037

出　　处：《林业科学》2023年第5期61-70,共10页Scientia Silvae Sinicae

基　　金：国家自然科学基金项目(32171790);江苏省现代农机装备与技术示范推广项目(NJ2020-18);江苏省研究生培养创新工程(SJCX21_0334)。

摘　　要：【目的】构建多视角植物表型信息采集平台,以固定时间间隔获取水肥胁迫处理下苗木图像,从形态变化中提取表征簸箕柳苗木水肥胁迫的应激反应表型参数,为苗木栽培和精确水肥管理提供参考。【方法】构建一套多视角表型信息采集平台,探讨一种基于时间序列图像的苗木表型信息提取方法,采用近似拼接算法将局部植株图像有效拼接成完整植株图像,基于多视角图像提出株高、主茎直径等表型信息的提取算法及株高可视化算法。以簸箕柳为研究对象开展水肥胁迫试验,设置水分因素和氮肥因素,包括正常水分处理与干旱胁迫处理2个水平以及6个氮肥水平,对比观察时间序列下簸箕柳株高和主茎直径的变化情况。【结果】1)图像近似拼接算法相较于传统直接拼接算法和暴力拼接算法在多视角表型信息采集平台上具有更好适应性,能够很好消除拼接痕迹,实现图像无缝拼接。2)多视角下株高提取算法的决定系数(R^(2))为0.96、主茎直径提取算法的R^(2)为0.74,能够解决单视角下因植物叶片遮挡导致的株高测量不准确和主茎直径无法测量等问题。3)整个生长周期时间序列下,簸箕柳苗木平均株高和主茎直径随时间的变化受不同水分处理显著影响,正常水分处理簸箕柳苗木平均株高和主茎直径分别为33.6 cm和2.49 mm,干旱胁迫处理簸箕柳苗木平均株高和主茎直径分别为22.8 cm和2.25 mm,正常水分处理簸箕柳苗木平均株高和主茎直径相比干旱胁迫处理高47.4%和10.7%。4)整个生长周期时间序列下,簸箕柳苗木平均株高和主茎直径在6个氮肥水平呈现出显著差异性,随氮肥含量增加呈先增加后降低趋势。正常水分处理下,除T5处理组外,其他组平均株高和主茎直径均高于CK处理组,其中T2处理组生长速率最高,速率分别为1.92 cm·d^(-1)和0.075 mm·d^(-1);干旱胁迫处理下,T1、T2处理组平均株高【Objective】Constructing a multi-view plant phenotypic information collection platform,image data of seedlings under water and fertilizer stress were obtained at fixed time intervals,and phenotypic parameters were extracted from morphological changes to characterize the stress response to water and fertilizer stress of Salix suchowensis seedlings,providing reference for cultivating seedling and managing water and fertilizer accurately.【Method】This paper constructed a multiple view phenotypic information collection platform and proposed a seedling phenotypic extraction method based on time sequence images by adopting the approximate stitching algorithm to realize the local image effective joining together into a complete image,phenotypic information such as plant height and major branch diameter extraction algorithm and plant height visualization algorithm were put forward based on image multiple points of view.In this paper,a water-fertilizer experiment was carried out with Salix suchowensis as the research object,and water and nitrogen factors were set up,including two levels of normal water treatment and drought stress treatment,and six levels of nitrogen concentration.The changes of plant height and main branch diameter of Salix suchowensis under time series were compared and observed.【Result】1)Compared with the traditional direct mosaic algorithm and violent mosaic algorithm,the approximate image mosaic algorithm proposed in this paper has better adaptability on the multi-perspective phenotypic information collection platform,and can eliminate the stitching traces well and realize the seamless image mosaic.2)The deterministic coefficient R^(2)of the plant height algorithm proposed in this paper is 0.96 under multi-view images,and that of the main branch diameter extraction algorithm under multi-perspective images is 0.74,which solves the problems of inaccurate measurement of plant height and un measurement of main branch diameter caused by plant leaf occlusion under single-perspective conditions.3)

关 键 词：表型信息 时间序列图像 株高 主茎直径 水肥胁迫 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]