中文摘要：

针对油菜水分胁迫的无损探测,研究了利用冠层反射光谱、多光谱图像和冠层温度多信息融合对油菜含水率进行定量分析的方法.采用逐步回归法对含水率的多传感特征进行提取,通过测量分析冠-气温差,结合环境温湿度信息,获取了水分胁迫指数（CWSI）特征,并对光强影响进行了特征补偿.研究发现960,1 450,1 650 nm的光谱特征,560 nm处的可见光图像,960,810 nm处的近红外图像均值及960,810 nm图像比值特征与油菜含水率的相关性均较高.利用主成分分析法（PCA）对特征空间进行变换和降维,进而BP神经网络建立了油菜含水率的多传感特征预测模型.结果表明,该方法能够利用多信息的综合作用优势实现对油菜水分胁迫的定量分析,模型精度与单一检测方法相比有显著提高.

英文摘要：

The canopy spectral reflectance,the multi-spectral image and the canopy temperature were fused to quantitatively analyze the rape moisture content based on nondestructive testing of rape water stress. Stepwise regression method was used to extract the features of moisture content from different sensors. The water stress index（ CWSI） was obtained by detecting canopy-air temperature difference and environment temperature and humidity to compensate light influence. The results show that the spectral features at wavelength of 960,1 450,1 650 nm,the features of image mean value at 560,960,810 nm and the image ratios at 960 nm to 810 nm are highly correlated with the rape moisture content during the whole growth period of rope. The principal component analysis（ PCA） was applied to transform and reduce dimensions for feature space,and the prediction model of moisture content of rape was built by BP neuralnetwork. The results show that more information can be integrated to achieve the quantitative analysis of water stress of rape. The proposed model precision is obviously higher than that of single detection method.