基于多源遥感数据的油菜长势监测

发布时间：2018-01-11 03:06

本文关键词：基于多源遥感数据的油菜长势监测　出处：《华中农业大学》2016年硕士论文　论文类型：学位论文

【摘要】：油菜是我国广泛种植的油料作物,实时、精准、快速估测油菜长势对于油菜生长诊断与管理以及产量预测具有重要意义。随着遥感技术不断发展,遥感数据源越来越丰富,使得大范围实时监测农作物长势成为可能,大量改进型光谱特征参数、植被指数反演农学参数为精准农业提供技术支持,尤其是高光谱遥感凭借波段连续性强、光谱信息量大的优势,为实时快速大面积监测作物长势提供有效信息。充分利用多源遥感数据监测油菜关键生长期的长势,建立较为精确的油菜生长遥感监测模型并提取油菜种植面积,成为大范围监测油菜长势监测的趋势。本研究优化了不同条件下(栽种方式、施肥水平等)油菜叶面积和叶鲜重的估算模型,采用高光谱技术反演了试验小区不同时期叶面积指数,并利用试验小区不同时期叶面积指数建立了试验小区油菜估产模型,探索通过提取遥感参数反演油菜叶面积指数,然后利用叶面积指数进行大范围估算油菜单产。上述研究得出结论如下:1.采用麦夸特法+通用全局优化算法对长宽系数模型的参数精确计算,建立的受外界干扰较少(栽种方式、施肥水平、生长时期)长宽幂函数模型,其建模效果和预测精度比常规长宽线性模型更高,可准确估计油菜叶片叶面积和叶鲜重。2.不同时期所建反演LAI模型以二次多项式为主,但最优模型差异较大,在苗期时以红边参数为主的建模精度和预测效果较好,而油菜生长后期(花期、角果期)则是以NLI为代表的非线性指数建模精度和预测效果相对较高,全生育所建模型精度较低,难以采用相对固定的光谱参数和指数来预测油菜整个生育期LAI,不同生育期的油菜株型和覆盖度存在较大的差异,花和角果与叶不同的光谱响应特征,采用统一建模预测LAI不能准确预测LAI,不同生长时期需要选择合适的光谱参数和植被指数建立预测模型。3.将不同时期试验小区叶面积指数作为自变量,试验小区最终产量作为因变量,相关分析表明,产量与各个时期LAI皆呈显著的正相关关系,与角果期叶面积指数呈现极显著正相关,通过逐步回归分析,利用十叶期、盛花期、角果期的叶面积指数所建模型精度较高,经检验具有较好地预测效果。4.利用遥感影像提取的9个常用植被指数逐步回归分别反演武穴市十叶期、盛花期、角果期的叶面积指数,利用试验小区叶面积指数—产量模型对武穴市油菜单产进行估计,结果表明利用遥感影像提取的参数可以有效估算三个时期油菜叶面积指数和油菜单产。
[Abstract]:Rapeseed is a widely grown oil crop in China. The real-time accurate and rapid estimation of rapeseed growth is of great significance for rape growth diagnosis management and yield prediction. With the development of remote sensing technology. Remote sensing data sources are more and more abundant, which makes it possible to monitor crop growth on a large scale in real time. A large number of improved spectral characteristic parameters and agronomic parameters of vegetation index inversion provide technical support for precision agriculture. In particular, hyperspectral remote sensing has the advantages of strong band continuity and large spectral information. It can provide effective information for real-time and rapid monitoring of crop growth and make full use of multi-source remote sensing data to monitor the growth of rapeseed in key growing period. The establishment of a more accurate remote sensing monitoring model of rape growth and the extraction of rape planting area have become the trend of monitoring rapeseed growth on a large scale. The leaf area and fresh weight of rape were estimated by using hyperspectral technique. The yield estimation model of rape was established by using the leaf area index of different periods in the experimental plot, and the retrieval of rape leaf area index by extracting remote sensing parameters was explored. Then using leaf area index to estimate rapeseed yield on a large scale. The conclusions are as follows: 1. The parameters of the length and width coefficient model are calculated accurately by using the general global optimization algorithm of McQuat method. The power function model with less external interference (planting mode, fertilization level, growth period) was established, and its modeling effect and prediction accuracy were higher than that of the conventional length and width linear model. The leaf area and fresh weight of rape leaves can be estimated accurately. The inverse LAI model built in different periods is mainly quadratic polynomial, but the difference of the optimal model is great. In seedling stage, the modeling accuracy and prediction effect of red edge parameter were better, while the late growth stage (flowering period, pod stage) of rapeseed was the nonlinear exponential modeling accuracy and prediction effect represented by NLI. The precision of the whole growth model is low, it is difficult to use the relative fixed spectral parameters and indices to predict the whole growth period of rape Lai. There are great differences in plant type and coverage of rape at different growth stages. The spectral response characteristics of flower, pod and leaf were different, and LAI could not be predicted accurately by using unified modeling to predict LAI. Different growth stages need to select the appropriate spectral parameters and vegetation index to establish a prediction model .3. take the leaf area index of different periods as the independent variable and the final yield of the experimental plot as the dependent variable. Correlation analysis showed that there was a significant positive correlation between yield and LAI in each stage, and a very significant positive correlation between yield and leaf area index in pod stage. Through stepwise regression analysis, the ten-leaf stage and flowering stage were used. The accuracy of the model established by the leaf area index of the pod period is high, and it has good prediction effect by testing. 4. Nine common vegetation indices extracted from remote sensing image were used to invert the ten leaf period and the full flowering stage of Wu-acupoint City respectively by stepwise regression. The leaf area index of the pod stage was estimated by using the leaf area index-yield model of the experimental plot to estimate the yield per unit of rape in Wucao city. The results showed that the parameters extracted from remote sensing images could effectively estimate rape leaf area index and rape yield in three periods.
【学位授予单位】：华中农业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：S565.4;S127

【参考文献】