摘要
研究利用遥感手段对棉花黄萎病不同严重度进行估测的可行性,为大面积监测棉花黄萎病提供理论依据。以新疆石河子市惠远种业大丽轮枝菌黄萎病病圃试验田为研究区,利用地物光谱仪分别采集的2019和2020年棉花冠层光谱反射率数据,结合棉花黄萎病严重度进行相关分析,建立基于偏最小二乘回归的病害严重度估测模型。对原始数据进行一阶微分、二阶微分、去包络线处理等数学变换,分析棉花黄萎病不同严重度之间的光谱特征,得到相应光谱特征波段,建立基于偏最小二乘回归的棉花黄萎病严重度估测模型。结果表明,不同黄萎病严重度的棉花在近红外及可见光波段表现不同特征;基于去除包络线后的光谱反射率数据与病害严重度之间的相关性最高,380~780 nm为去除包络线光谱病害敏感波段区间,其中380~600 nm、615~695 nm波段与病害严重度相关系数达到0.8,达到极显著水平(P<0.01),为棉花黄萎病病害严重度敏感波段;构建的基于去包络线光谱数据的偏最小二乘回归估测模型R^(2)达到0.764,为病害严重度估测的最佳模型。
To study the feasibility of estimating different severity of cotton verticillium wilt by remote sensing with a view to provide theoretical basis for monitoring cotton verticillium wilt in large area.The experimental field of verticillium dahliae verticillium wilt nursery in Huiyuan Seed Industry Farm of Shihezi City,Xinjiang was the study area,the spectral reflectance data of cotton canopy in 2019 and 2020 were collected by spectrometer,and the correlation analysis was carried out with the severity of cotton verticillium wilt,and the disease severity estimation model based on partial least squares regression was established.First-order differential,second-order differential,de-enveloping processing and other mathematical transformations were carried out on the original data,and the spectral characteristics of different severities of cotton verticillium wilt were analyzed,and the corresponding spectral characteristic bands were obtained,and the cotton verticillium wilt severity estimation model based on partial least squares regression was established.The results showed that cotton with different severities of cotton verticillium wilt showed different characteristics in near infrared and visible light bands;The correlation between spectral reflectance data after envelope removal and disease severity was the highest,380-780 nm was the spectral sensitive band of cotton verticillium wilt,and the correlation coefficient between 380-600 nm and 615-695 nm and disease severity was 0.8,which was extremely significant(P<0.01).The constructed partial least squares regression estimation model R^(2) based on de-enveloping spectral data reached 0.764,which was the best model for disease severity estimation.
作者
陈巧玲
范燕敏
武红旗
杨强军
王德俊
邓亚辉
王舂霞
CHEN Qiao-ling;FAN Yan-min;WU Hong-qi;YANG Qiang-jun;WANG De-jun;DENG Ya-hui;WANG Chun-xia(l.College of Pratacultural and Environmental Sciences,Xinjiang Agricultural University,Urumqi 830052,China;College of Agronomy,Xinjiang Agricultural University,Urumqi 830052,China)
出处
《新疆农业大学学报》
CAS
2020年第4期261-269,共9页
Journal of Xinjiang Agricultural University
基金
地区科学基金项目(31560340)。
关键词
高光谱
棉花黄萎病
黄萎病严重度
估测模型
偏最小二乘回归
hyperspectral
verticillium wilt of cotton
severity of verticillium wilt
estimation model
partial least squares regression
