【作者】 张畅；

【导师】 方慧；

【作者基本信息】 浙江大学 ， 农业电气化与自动化， 2015， 硕士

【摘要】 应用遥感技术和计算机技术获取植物的生长信息、监测植物的生长状况,已成为现代精细农业领域中一个重要环节。叶片是植物进行光合作用的重要器官,植物对水分、养分的吸收与叶片结构、生物量、色素等密切相关。叶片光学特性的表征受水分、养分利用率,环境胁迫下的自适应等外界条件的影响。因此,测定叶片半球的光学特性,准确、高效地表征作物生长参数,进行定量分析,在植被遥感研究与应用中有着极大的优势。本文以玉米叶片作为试验对象,应用光学原理和模型,基于自主研发光学特性采集平台,优化了试验条件。重点研究了不同背景下玉米叶片反射光谱的扣除方法,利用漫反射标准板的分布规律校正了试验误差,实现了氮素胁迫下玉米叶片多角度多方位半球反射分布特性的模拟。研究的主要成果有：(1)分析背景因素对玉米叶片光学特性的影响。基于4点假设,首次提出了BPLT (Background-Plate)模型。模型的最终表达式为R=R12+(1-R12)*(1-R1)*t2*(1+R221*t2)*R21。在背景材料中选取3种背景作为一组输入,对不同叶绿素浓度的60组玉米叶片进行同样的4组平行处理。：当确定系数(DC)大于0.68,误差离差平方和(SSE)小于7.5的玉米叶片背景处理组适用于BPLT模型。较高叶绿素浓度对背景扣除效果好,稳定性高；4种平行处理,尤以T-K-B背景组(Rb)扣除的效果最为稳定。结合叶绿素指数、方差分析法对模型进行验证,优选5种叶绿素指数NDI,NDVI,SAVI,OSAVI,ARVI,作为衡量背景单因素BPLT模型扣除效果的指标；优选指数PSSR作为衡量背景、叶绿素浓度二因素BPLT模型扣除效果的指标。(2)基于Spectralon板(S板)近似Lambertian体的光学特性,针对实验室自主设计的植物三维漫反射辐射仪平台进行校正。以光学模型双向反射分布函数(BRDF,fr)、定向半球反射因子(DHRF,p)为基础,将试验采集数据转换成fr最终表达式的一部分。确定了探测器最佳天顶角θr,degree=30°,其残差平方和(Rss)平均值为0.046。建立了不同光源入射角下的探测器天顶角(X, rad)与比例系数(Y)的余弦函数关系式：Y=1.3938cosX-0.2127.简化了fr,leaf的最终表达式：上式的推导,避免了光纤扰动的影响,减少实验室重复测量次数,极大缩短了S板校正时间。(3)研究氮素胁迫下玉米叶片双向反射分布函数fr,leaf的变化规律。研发了基于Matlab GUI界面的玉米叶片BRDF光学特性数据处理平台,主要包括数据导入、数据的存放、BRDF的计算、绘图四个部分,提高数据批量处理的效率。绘制不同波段、不同光源位置、不同方位、不同氮素胁迫的玉米叶片双向反射分布函数图,并以各向异性因子(ANIX)进行评价。结果表明,随光源角度的增加,氮素胁迫下玉米叶片的BRDF前向散射效应明显,光源入射角θi=45°时为本试验研究玉米氮素胁迫的理想条件,在680 nm波段更容易分析玉米叶片双向反射分布函数图的分布规律,不同氮素胁迫下,玉米叶片双向反射分布函数趋势相同,但中心位置处亮度不同,光源入射到玉米叶片间叶中位置处时,更能反映不同样本间光学性质的差异。更多还原

【Abstract】 One of major part in precision agriculture is the acquisition information of plant growth and the diagnosis of the growth condition applied by remote sensing technology and computer graphics. In addition, water or nutrient absorption on plants is closely related to leaf structure, biomass, pigment etc., leaf is an indicative organ for plant photosynthesis. Measurement of optical properties varied with external conditions, such as nutrient use efficiency, adaption to environmental stress, is of promising prospect. Therefore, characterization of leaf optical properties, including accurate and efficient measurement of indices of maize (Zea mays) leaves and the quantitative analysis methods meets the needs in remote sensing and application of crops. In this thesis, using maize (Zea mays) leaves as research object, optical principle and model were applied to optimize the experiment conditions of the self-developed optical acquisition system. Background elimination for maize (Zea mays) leaves based on the BPLT model was established, experimental error was reduced by the correction of the distribution of Spectralon. Moreover, light reflectance distribution compared with multi-angle of nitrogen stress or water stress measurements in maize leaves was completed. The main contents and brief conclusion are as follows:(1) In order to precisely acquire leaf reflectance spectra, influence of background on leaf reflectance spectra was studied. Experiment was conducted to discriminate the characteristics of 60 maize (Zeamays) leaves based on 8 background materials and leaf chlorophyll concentration. BPLT (Background Plate) model, which premised on a set of assumptions,R=R12+(1-R12)*(1-R21)*t2*(1+R221*t2)*R21, was promoted and applied to remove the influence of background material. To verify this model, Analysis of Variance (ANOVA) was conducted. The results indicated that maize (Zea mays) leaf could be effectively estimated by the means of BPLT model with determined coefficients (DC) greater than 0.68 and residual sum of squares (SSE) less than 7.5. Meanwhile, leaf chlorophyll concentration at high levels was stable with the T-K-B background. As with the ANOVA, vegetation indices NDI, NDVI, SAVI, OSAVI, ARVI were better by One-way ANOVA while vegetation indices PSSR was better to study the influence of different chlorophyll concentration.(2) The preference of the homogeneity of the reference Spectralon panel was required to be calibrated to eliminate the corresponding factor. Bidirectional reflectance distribution factor (BRDF,fr), directional hemispherical reflectance factor (DHRF, p) as well as the measured data was held on to the conduction of fr,leaf. It was considered sensor zenith angle θr,degree=30°, of which was the lowest average root sum square (Rss=0.046), as the standard angle to decrease the error combined with fitting. Coefficient factor (Y) was plotted as a function of θr (in radiance, X) of different illumination angles Y=1.3938cosX-0.2127 (θi=0°～45°). The ultimate mathematical expression of fr,leaf was Thus, the error suffered from the curvature of the optical fiber or redundancy of repeated measurement data was reduced.(3) Light reflectance distribution compared with multi-angle of nitrogen stress measurements on maize (Zea mays) leaves was conducted based on the above study. Matlab GUI interface was developed for batch processing of optical data of maize (Zea mays) leaves. It was divided into four parts:data importing, data retention, calculation of fr,leaf and figure. Concerning different wavelength, different zenith/azimuth angle of nitrogen stress on maize (Zea mays) leaves, polar plot added with ANIX (Anistrophy Index) was demonstrated. Results showed that the maximum reflectance peak of maize (Zea mays) leaves was out of the principal plane besides the specular direction within increasing illumination zenith angle, in that case, the ideal experiment condition could be at θi=45° while wavelength of 680nm was more adaptable to analyze the characteristic of its distributing law. Moreover, the middle part of maize (Zea mays) leaves interacted with different illumination zenith angles were better to compare the optical properties among themselves.更多还原