基于高光谱成像技术的土壤水盐及番茄植株水分诊断机理与模型研究

发布时间：2018-09-13 14:25

【摘要】：宁夏回族自治区地处我国西部黄河上游,属典型的大陆性半湿润半干旱气候。我区的特色经济作物产业,是宁夏地区最有发展潜力的农业增收项目之一,而土壤水肥的精确补给直接影响作物高产、优质。因此,如何采取廉价、快速、省力的手段来获取干旱区和半干旱区地域范围内大面积盐渍化土壤盐水分的动态信息,对盐渍化土壤的治理、合理规划利用具有重要意义。本研究以温室番茄植株为研究对象,基于Vis-NIR与NIR高光谱成像技术,结合化学计量学方法,对土壤水分、盐分以及番茄植株水分动态监测,为快速诊断植株水分亏缺程度以及土壤水盐分检测机理研究提供理论依据。主要研究结果包括:(1)不同灌水条件下,土柱中的土壤含水率变化情况不同。对不同含盐量土壤来说,土壤中盐分的含量对土壤水分的再分布具有较大的影响;相比2%浓度含盐量灌溉的土柱,0.2%浓度含盐量灌溉的土柱可以更好的控制水分在土壤中的运移。对土壤的不同含水率、含盐量变化规律进行了分析,建立了4种表层土壤与深层土壤水分的数学模型。(2) 土壤的反射率随着土壤含水率的增加而减小,当增大到超过田间持水率时,土壤的反射率会随着土壤含水率的增加而增大。并探讨了土壤水分的不同方法提取特征波长、不同建模方法、不同光谱范围及特征波长与全波段的建模效果,优选900～1700nm波段建立的SPA方法提取特征波长的MLR模型,筛选出特征波长为987、1386、1466、1568、1636、1645 nm, 土壤含水率最优模型的预测相关系数(Rp)为0.984,预测均方根误差(RMSEP)为0.631。(3)随着土壤中含盐量的增加,土壤中水分蒸发情况受到的影响不同;不同天数下,不同波段下体现了随着土壤含盐量的增加,土壤光谱的反射率增大;而对于高含盐量土壤,土壤反射率变化差异较小。这为今后智能遥感定性判别土壤含盐量提供理论依据。(4)探讨了土壤盐分的不同方法提取特征波长、不同建模方法、不同光谱范围及特征波长与全波段的建模效果,优选900～1700nm波段的β系数方法提取特征波长的PLSR模型,筛选出特征波长为 936、996、1016、1136、1151、1186、1273、1395、1425、1458、1535、1642nm,土壤含盐量的预测相关系数(Rp)为0.949,预测均方根误差(RMSEP)为2.914g/kg。(5)研究了番茄叶片的光谱信息与水分含量直接的关系以及盐-水耦合的生物控制机制。探讨了番茄叶片的不同方法提取特征波长、不同建模方法、不同光谱范围及特征波长与全波段的建模效果,优选900～1700nm波段的SPA提取特征波长的PLSR模型,筛选出特征波长为918、981、1029、1387、1652nm,叶片的含水率的预测相关系数(Rp)为0.9,预测均方根误差(RMSEP)为 0.614。(6)利用高光谱成像技术对土壤的水分、盐分以及温室番茄植株的水分进行模型构建,将深层土壤、表层土壤、番茄冠层与高光谱建立联系,为宁夏区域土壤水盐含量遥感与植物叶片水分快速检测奠定基础。
[Abstract]:Ningxia Hui Autonomous region is located in the upper reaches of the Yellow River in western China, which is a typical continental semi-humid and semi-arid climate. The characteristic cash crop industry in our region is one of the most potential agricultural income increasing projects in Ningxia, and the accurate supply of soil water and fertilizer directly affects the high yield and high quality of crops. Therefore, how to use cheap, rapid and labor-saving means to obtain the dynamic information of salinized soil salt water distribution in arid and semi-arid areas is of great significance to the treatment of salinized soil and rational planning and utilization of salinized soil. Based on Vis-NIR and NIR hyperspectral imaging technique and chemometrics method, the dynamic monitoring of soil moisture, salt content and tomato water content in greenhouse tomato plants was studied. It provides a theoretical basis for the rapid diagnosis of water deficit in plants and the study on the mechanism of soil water salinity detection. The main results are as follows: (1) the variation of soil moisture content in soil column is different under different irrigation conditions. For different salinity soil, the salt content in the soil has a greater impact on the redistribution of soil moisture, compared with the soil column with 2% salinity irrigation, the soil column with 0.2% salt concentration irrigation can better control the movement of water in the soil. The variation law of soil moisture content and salt content was analyzed, and four mathematical models of surface soil and deep soil moisture were established. (2) soil reflectivity decreased with the increase of soil moisture content. The soil reflectivity increases with the increase of soil moisture content when the field water holdup is increased. Different methods of extracting characteristic wavelength, different modeling methods, different spectral range, characteristic wavelength and the modeling effect of the whole wave band were discussed. The MLR model of extracting characteristic wavelength by SPA method in 900~1700nm band was selected. The predicted correlation coefficient (Rp) and root mean square error (RMSEP) of the optimal model of soil moisture content of 987 ~ 1386N 146N 1568336 ~ 1645 nm, were 0.984 and 0.631respectively. (3) with the increase of salt content in the soil, the evaporation of soil water was affected by different days, and the correlation coefficient was 0.984, and the root mean square error (RMSEP) was 0.631. (3) with the increase of salt content in the soil, the evaporation of soil water was affected by different days. The spectral reflectance of soil increased with the increase of soil salt content in different bands, but the variation of soil reflectivity was small for high salinity soil. This provides a theoretical basis for intelligent remote sensing to qualitatively judge soil salinity. (4) the modeling effects of different methods for extracting soil salinity, different modeling methods, different spectral ranges, characteristic wavelengths and full wavelengths are discussed. The PLSR model of characteristic wavelength was extracted by 尾 -coefficient method in 900~1700nm band. The characteristic wavelength was 936 / 9961016 / 11363 / 1151 / 11866 / 12773 / 1395N / 1425 / 1455 / 1535N / 1642nm, the predicted correlation coefficient of soil salt content was 0.949 and the (RMSEP) of predicting root mean square error was 2.914g / kg / g. (5) the direct relationship between spectral information and water content in tomato leaves and the biological control mechanism of salt-water coupling were studied. Different methods of extracting characteristic wavelengths, different modeling methods, different spectral ranges, characteristic wavelengths and full-band modeling effects of tomato leaves were discussed. The PLSR model of extracting characteristic wavelengths of SPA in 900~1700nm band was selected. The characteristic wavelength was 918 ~ 981 ~ 1029 ~ 13877N ~ (1652) nm, the predicted correlation coefficient of water content in leaves was 0.9and the root mean square error (RMSEP) of prediction was 0.614. (6) the model of soil moisture, salt content and tomato water in greenhouse was constructed by hyperspectral imaging technique. The deep soil, surface soil and tomato canopy were linked with hyperspectral data, which laid a foundation for remote sensing of soil water and salt content in Ningxia region and rapid detection of water content in plant leaves.
【学位授予单位】：宁夏大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：S156.4;S641.2

【参考文献】