基于高光谱遥感的水稻生长监测研究
发布时间：2018-08-01 11:31
【摘要】：快速、实时、准确的获取农田生态环境和作物长势信息是精准农业实施的重要基础前提,也是现代精准农业发展的关键技术瓶颈之一。本论文把水稻作为研究对象,以不同施肥(氮肥、生物质炭肥)梯度下的小区试验作为依托,综合运用高光谱遥感、生理生化参数测试以及数理统计等技术手段,分析不同施肥(氮肥、生物质炭肥)条件下水稻在不同生育期的冠层高光谱特征、叶绿素含量(SPAD)特征和叶面积指数(LAI)特征,分生育期的形式建立基于原始光谱反射率参数、“三边”参数的水稻叶绿素含量、叶面积指数的高光谱估算模型,并利用决定系数(R2)、均方根误差(RMSE)、相对误差等评价指标验证预测模型的精度。本研究主要取得了以下结果:(1)对水稻在不同生育期以及不同施肥(氮肥、生物质炭肥)条件下的冠层光谱的变化规律进行研究。结果表明:从拔节期到乳熟期,水稻冠层光谱反射率在可见光区域是不断增大的,而在近红外区域是先增大后减小的。随着施氮水平的提高,水稻冠层光谱反射率在可见光范围内总体是呈现逐渐降低的趋势,而在近红外波段是逐渐增高的。不同生物质炭水平下,水稻在可见光波段的冠层光谱的反射率差别不明显,而到达近红外波段,冠层光谱反射率的差异较明显并且施炭处理的水稻冠层光谱反射率的值大于不施炭处理的冠层光谱反射率的值。因此,该研究结果能够为更深层次地利用水稻冠层光谱信息监测水稻生长状况提供一定的理论基础。(2)分析了水稻冠层原始光谱、导数光谱与叶绿素含量(SPAD)、叶面积指数(LAI)的相关关系。在可见光区域,水稻原始光谱反射率与叶绿素含量、叶面积指数在拔节期、抽穗期与灌浆期呈现出负相关,在“红边”处,由负相关变成正相关;水稻的导数光谱与叶绿素含量、叶面积指数之间的相关系数在拔节期、抽穗期与灌浆期的一些波段处高于原始光谱反射率与叶绿素含量、叶面积指数的相关系数;乳熟期水稻的原始光谱、导数光谱与叶绿素含量、叶面积指数的相关系数较低,因此,在建立水稻叶绿素含量、叶面积指数的高光谱估算模型过程中,不适宜使用乳熟期的光谱数据。(3)应用2014年的水稻冠层光谱与叶绿素含量(SPAD)、叶面积指数(LAI)数据,在水稻拔节期、抽穗期与灌浆期基于原始光谱反射率参数(绿峰位置、绿峰反射率、绿峰面积、红谷位置、红谷反射率、红谷面积、绿峰面积与红谷面积的比值与归一化值)、“三边”参数(蓝边位置、蓝边振幅、蓝边面积、黄边位置、黄边振幅、黄边面积、红边位置、红边振幅、红边面积、红边面积与蓝边面积的比值与归一化值、红边面积与黄边面积的比值与归一化值)建立叶绿素含量、叶面积指数的估算模型,并应用2013年水稻冠层光谱与叶绿素含量、叶面积指数的相关数据对所建的预测模型进行精度检验。结果表明:运用原始光谱反射率参数反演水稻叶绿素含量时,在拔节期应优先考虑红谷面积,抽穗期、灌浆期的红谷反射率、红谷面积反演效果都很好。使用原始光谱反射率参数反演水稻叶面积指数时,拔节期,优先考虑绿峰面积与红谷面积的归一化值,抽穗期,红谷反射率、红谷面积的反演效果相对较好,灌浆期的绿峰反射率、红谷面积反演效果相对较好。应用“三边”参数反演水稻叶绿素含量时,拔节期优先考虑应用红边面积与蓝边面积的比值,抽穗期优先考虑使用蓝边面积,灌浆期红边位置、红边面积与蓝边面积的归一化值反演效果都较好。使用“三边”参数反演水稻叶面积指数时,拔节期,红边面积、红边面积与蓝边面积的归一化值的反演效果都很好,抽穗期,优先考虑红边面积,灌浆期,红边位置、红边面积与蓝边面积的比值的反演效果相对较好。
[Abstract]:Fast, real-time, accurate acquisition of farmland ecological environment and crop growth information is an important basic prerequisite for the implementation of precision agriculture. It is also one of the key technology bottlenecks in the development of modern precision agriculture. This paper takes rice as the research object and relies on the plot experiment under the gradient of nitrogen fertilizer (nitrogen fertilizer, biomass charcoal fertilizer). The spectral remote sensing, physiological and biochemical parameters and mathematical statistics were used to analyze the hyper spectral characteristics, chlorophyll content (SPAD) characteristics and leaf area index (LAI) characteristics of rice at different growth stages under different fertilization (nitrogen fertilizer, biomass carbon fertilizer), based on the original spectral reflectance parameters, "three sides". The chlorophyll content of rice, the Hyperspectral Estimation Model of leaf area index, and the evaluation index of R2, RMSE and relative error were used to verify the accuracy of the prediction model. The main results are as follows: (1) under the condition of different growth period and different fertilization (nitrogen fertilizer, biomass carbon fertilizer) The spectral reflectance of the canopy was studied. The results showed that the spectral reflectance of the rice canopy increased from the jointing stage to the milk ripening period, but increased first and then decreased in the near infrared region. With the increase of nitrogen application level, the spectral reflectance of rice canopy was gradually decreasing in the visible light range. In the near infrared band, the reflectance of the canopy spectral reflectance in the visible light band is not obvious, but the reflectance of the canopy spectral reflectance is obvious in the near infrared band, and the value of the canopy spectral reflectance of the rice is greater than that of the canopy spectral reflectance without carbon treatment. Therefore, the results can provide a theoretical basis for monitoring rice growth by using the spectral information of rice canopy more deeply. (2) the correlation between the original spectrum of the rice canopy, the derivative spectrum and the chlorophyll content (SPAD), the leaf area index (LAI), and the original spectral reflectance of rice in the visible region, and the reflectance of the rice were analyzed. The content of chlorophyll, leaf area index at the jointing stage, the heading stage and the filling stage showed a negative correlation, in the "red edge", from negative correlation into positive correlation, the correlation coefficient between the derivative spectrum of the rice and the chlorophyll content, the leaf area index was at the jointing stage, and some bands at the heading and filling stages were higher than those of the original spectral reflectance and chlorophyll. The correlation coefficient of the content, leaf area index, the original spectrum of rice, the derivative spectrum and the chlorophyll content and the leaf area index are lower. Therefore, in the process of establishing the Hyperspectral Estimation Model of the chlorophyll content and leaf area index of rice, the spectral data of the milk ripening period are not suitable. (3) the application of the rice canopy light in 2014. The spectrum and chlorophyll content (SPAD) and leaf area index (LAI) data are based on the original spectral reflectance parameters (green peak position, green peak reflectivity, green peak area, Red Valley location, Red Valley area, Red Valley area, green peak area and Red Valley area ratio and normalization value), and "three edge" parameter (blue edge position). The amplitude of blue edge, blue edge area, yellow edge position, yellow edge amplitude, yellow edge area, red edge position, red edge amplitude, red edge area, red edge area and blue edge area ratio and normalized value, ratio and normalization value of red edge area and yellow edge area and normalization value) set up chlorophyll content, estimation model of leaf area index, and applied the spectrum of rice canopy in 2013. The results showed that the Red Valley area, the heading stage, the Red Valley reflectivity at the grain filling period and the Red Valley area inversion effect were good. When the rice leaf area index was retrieved by the ejection parameter, the jointing period was given priority to the normalized value of green peak area and Red Valley area. The inversion effect of the heading stage, the Red Valley reflectivity and the Red Valley area was relatively better. The green peak reflectivity and the Red Valley area inversion effect was relatively better. When the chlorophyll content was retrieved with the "three edge" parameters, the extraction was extracted. In the festival, the ratio of red edge area to blue edge area was first considered, and the area of blue edge was first considered in heading stage, and the inversion effect of red edge area and blue edge area was better. When the rice leaf area index was retrieved by "three edge" parameters, the extraction period, red edge area, red edge area and blue edge area were returned. The inversion effect of one value is very good, and the back effect of the ratio of red edge area, red edge location, red edge area and blue edge area is better than that of red edge area.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S127;S511
本文编号：2157437
[Abstract]:Fast, real-time, accurate acquisition of farmland ecological environment and crop growth information is an important basic prerequisite for the implementation of precision agriculture. It is also one of the key technology bottlenecks in the development of modern precision agriculture. This paper takes rice as the research object and relies on the plot experiment under the gradient of nitrogen fertilizer (nitrogen fertilizer, biomass charcoal fertilizer). The spectral remote sensing, physiological and biochemical parameters and mathematical statistics were used to analyze the hyper spectral characteristics, chlorophyll content (SPAD) characteristics and leaf area index (LAI) characteristics of rice at different growth stages under different fertilization (nitrogen fertilizer, biomass carbon fertilizer), based on the original spectral reflectance parameters, "three sides". The chlorophyll content of rice, the Hyperspectral Estimation Model of leaf area index, and the evaluation index of R2, RMSE and relative error were used to verify the accuracy of the prediction model. The main results are as follows: (1) under the condition of different growth period and different fertilization (nitrogen fertilizer, biomass carbon fertilizer) The spectral reflectance of the canopy was studied. The results showed that the spectral reflectance of the rice canopy increased from the jointing stage to the milk ripening period, but increased first and then decreased in the near infrared region. With the increase of nitrogen application level, the spectral reflectance of rice canopy was gradually decreasing in the visible light range. In the near infrared band, the reflectance of the canopy spectral reflectance in the visible light band is not obvious, but the reflectance of the canopy spectral reflectance is obvious in the near infrared band, and the value of the canopy spectral reflectance of the rice is greater than that of the canopy spectral reflectance without carbon treatment. Therefore, the results can provide a theoretical basis for monitoring rice growth by using the spectral information of rice canopy more deeply. (2) the correlation between the original spectrum of the rice canopy, the derivative spectrum and the chlorophyll content (SPAD), the leaf area index (LAI), and the original spectral reflectance of rice in the visible region, and the reflectance of the rice were analyzed. The content of chlorophyll, leaf area index at the jointing stage, the heading stage and the filling stage showed a negative correlation, in the "red edge", from negative correlation into positive correlation, the correlation coefficient between the derivative spectrum of the rice and the chlorophyll content, the leaf area index was at the jointing stage, and some bands at the heading and filling stages were higher than those of the original spectral reflectance and chlorophyll. The correlation coefficient of the content, leaf area index, the original spectrum of rice, the derivative spectrum and the chlorophyll content and the leaf area index are lower. Therefore, in the process of establishing the Hyperspectral Estimation Model of the chlorophyll content and leaf area index of rice, the spectral data of the milk ripening period are not suitable. (3) the application of the rice canopy light in 2014. The spectrum and chlorophyll content (SPAD) and leaf area index (LAI) data are based on the original spectral reflectance parameters (green peak position, green peak reflectivity, green peak area, Red Valley location, Red Valley area, Red Valley area, green peak area and Red Valley area ratio and normalization value), and "three edge" parameter (blue edge position). The amplitude of blue edge, blue edge area, yellow edge position, yellow edge amplitude, yellow edge area, red edge position, red edge amplitude, red edge area, red edge area and blue edge area ratio and normalized value, ratio and normalization value of red edge area and yellow edge area and normalization value) set up chlorophyll content, estimation model of leaf area index, and applied the spectrum of rice canopy in 2013. The results showed that the Red Valley area, the heading stage, the Red Valley reflectivity at the grain filling period and the Red Valley area inversion effect were good. When the rice leaf area index was retrieved by the ejection parameter, the jointing period was given priority to the normalized value of green peak area and Red Valley area. The inversion effect of the heading stage, the Red Valley reflectivity and the Red Valley area was relatively better. The green peak reflectivity and the Red Valley area inversion effect was relatively better. When the chlorophyll content was retrieved with the "three edge" parameters, the extraction was extracted. In the festival, the ratio of red edge area to blue edge area was first considered, and the area of blue edge was first considered in heading stage, and the inversion effect of red edge area and blue edge area was better. When the rice leaf area index was retrieved by "three edge" parameters, the extraction period, red edge area, red edge area and blue edge area were returned. The inversion effect of one value is very good, and the back effect of the ratio of red edge area, red edge location, red edge area and blue edge area is better than that of red edge area.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S127;S511
【共引文献】
相关期刊论文 前4条
1 李桂娟;朱丽丽;李井会;;作物氮素营养诊断的无损测试研究与应用现状[J];黑龙江农业科学;2008年04期
2 唐延林,王秀珍,王珂;利用光谱法测定水稻生物物理参数及其与光谱变量的相关性研究[J];贵州大学学报(农业与生物科学版);2002年05期
3 李晓利;唐延林;楼佳;;杂交水稻叶片叶绿素含量与光谱吸收率的关系[J];山地农业生物学报;2011年01期
4 唐延林,王人潮,王秀珍;对水稻微分光谱和植被指数的探讨[J];上海交通大学学报(农业科学版);2003年03期
相关硕士学位论文 前4条
1 张晓倩;基于全极化SAR数据的水稻识别与作物参数估算模型[D];中国地质大学(北京);2012年
2 李燕强;不同冬小麦品种冠层光谱特征及其与农学参数关系研究[D];河南农业大学;2012年
3 阎国倩;植被与土壤混合像元的高光谱偏振特性研究[D];东北师范大学;2013年
4 武婕;玉米成熟期地上生物量及其碳氮累积量的遥感估算[D];山东农业大学;2014年
，本文编号：2157437
本文链接：https://www.wllwen.com/kejilunwen/nykj/2157437.html
