低温胁迫下冬小麦生理活性的高光谱监测研究

发布时间：2018-06-19 04:33

本文选题：生理活性 + 低温胁迫　；参考：《山西农业大学》2015年硕士论文

【摘要】：为研究低温胁迫下冬小麦生理活性变化高光谱监测的可行性,本研究以晋太182(强冬性)和济22(半冬性)两个冬小麦品种为供试材料,采用盆栽设计,使用人工霜冻箱对拔节期冬小麦进行时间(4h,8h,12h)和温度(-2℃,-4℃,-6℃)两个因素共9个处理的低温胁迫。测定低温胁迫后不同生育时期功能叶过氧化物酶(POD)、超氧化物歧化酶(SOD)和丙二醛(MDA)的生理活性变化,并采集对应冬小麦冠层光谱,对光谱(全谱和特征光谱)和冬小麦生理活性(POD、SOD、MDA)进行多元线性回归分析,构建低温胁迫后冬小麦生理活性监测模型。结果表明：1.在冬小麦的整个生育期中,对照组冬小麦从拔节期到灌浆期,POD活性呈升高趋势,SOD和MDA没有明显变化,趋势比较平缓。低温胁迫后,冬小麦功能叶的POD活性变化总体趋势仍与对照组相同,但仍可以看出,强冬性品种的抗寒性较强,而半冬性品种恹复能力较强。超氧化物歧化酶保护系统对胁迫时间的敏感程度高于胁迫温度：MDA积累量没有表现出明显变化。2.从化学维分析,POD、SOD 和 MDA三种生理指标的活力值分布范围都很广并且验证集活力值均包含于建模集范围内。从平均值和标准差分析,建模集和验证集的偏移幅度也比较理想,说明所选建模集数据和验证集的数据划分具有代表性,同样数据量也满足建模的要求,可以进行建模。3.从光谱维分析,实现基于POD、SOD、MDA三种生理因子活性的监测是可行的,并且分析结果表明,SG-MLR模型建模效果最好,预测效果最佳。每个时期筛选出的特征波段,全部分散分布于可见光和近红外范围,但大部分位于可见光区域,特别是在400-450 nm波段范围更为集中。具体表现为,从拔节期到灌浆期80%以上的特征波段都位于可见光范围。除抽穗期外,50%以上的特征波段位于400-450 nm波段范围,即三种生理指标的活性变化在该波段范围携带有更多的有效信息。4.通过全波段粤特征波段建模的分析比较得知,特征波段建模(SPA-MLR)与全谱建模(SG-MLR)的建模集相关系数RC、RP及校正标准差(RMSEc),预测标准差(RMSEp)差异性不大(抽穗期差异较大),但是采用特征波段进行建模,数据量会减少95%左右,大大减少了共线和冗余信息,减轻了计算工作量,降低了模型复杂程度。因此,可以用特征波段代替全波段进行建模。
[Abstract]:In order to study the feasibility of hyperspectral monitoring of physiological activity changes of Winter Wheat under low temperature stress, this study used two winter wheat varieties of Jintai 182 (strong winter) and 22 (semi winter) as the test materials, using pot design, using artificial frost boxes for winter wheat at jointing stage (4h, 8h, 12h) and temperature (-2 C, -4, -6 C) 9 factors. The physiological activities of functional leaf peroxidase (POD), superoxide dismutase (SOD) and malondialdehyde (MDA) were measured at different growth stages after low temperature stress, and the canopy spectra of winter wheat were collected, and the spectral (total and characteristic spectra) and physiological activity of Winter Wheat (POD, SOD, MDA) were analyzed by multiple linear regression analysis, and the construction of winter wheat physiological activity (POD, SOD, MDA) was analyzed. The results of the physiological activity monitoring model of winter wheat after low temperature stress showed that: 1. during the whole growth period of winter wheat, the activity of POD in the control group increased from the jointing stage to the filling stage, and there was no obvious change in SOD and MDA, and the trend was relatively slow. The overall trend of POD activity in winter wheat was still in phase with the control group after low temperature stress. It is still clear that the cold resistance of the strong winter varieties is stronger and the semi winter variety is stronger. The sensitivity of the superoxide dismutase protection system to the stress time is higher than that of the stress temperature. The MDA accumulation does not show significant changes in the distribution of the activity values of the three physiological indexes of.2., POD, SOD and MDA. From the mean value and the standard deviation analysis, the offset of the modeling set and the validation set is also ideal. It shows that the data partition of the selected modeling set and the validation set is representative. The same amount of data also satisfies the requirements of the modeling. The modeling.3. can be analyzed from spectral dimension, and the model can be modeled. It is feasible to monitor the activity of three physiological factors based on POD, SOD and MDA. And the analysis results show that the SG-MLR model has the best modeling effect and the best prediction effect. All the characteristic bands selected at each time are distributed in the visible and near infrared range, but most of them are in the visible light region, especially at the 400-450 nm band range. The feature bands above 80% from the jointing period to the filling stage are in the visible light range. In addition to the heading period, more than 50% of the characteristic bands are located in the 400-450 nm band range, that is, the activity changes of three physiological indexes carry more effective information.4. through the full band Guangdong characteristic band modeling. The correlation coefficient RC, RP and correction standard deviation (RMSEc) of characteristic band modeling (SPA-MLR) and full spectrum modeling (SG-MLR) are analyzed and compared. The difference of prediction standard deviation (RMSEp) is not significant (the difference of heading period is large). However, the data amount will be reduced by about 95% with the feature band modeling, which greatly reduces the collinear and redundant information and reduces the calculation. Workload can reduce the complexity of the model, so the characteristic band can be used instead of the full band for modeling.
【学位授予单位】：山西农业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：S127;S512.11

【参考文献】