摘要
为寻求水稻被褐飞虱侵害后冠层温度特征的有效评估方法,以褐飞虱易感水稻品种“TN1”为研究对象,设置了褐飞虱侵害及未侵害两个处理,运用热红外成像技术获取水稻的冠层温度特征,使用机器学习分类器,对褐飞虱诱导的水稻冠层热图像温度特征变异评估方法进行了研究。首先,对试验采集的水稻冠层热图像和对应时刻的空气温度、相对湿度以及水稻灌溉水层水温信息进行分析,针对水稻冠层热图像提取了3种统计学温度特征,并使用了累计差值法分析水稻冠层的特征数;然后,对空气温度、相对湿度、水温与冠层温度特征分别进行了相关性分析;最后,分别采用逻辑回归算法与支持向量机算法进行评估模型的拟合。结果表明:3种统计学特征中,冠层温度变异系数的累计差值为30.78,是差异性最大的特征值;统计学特征与空气温度、相对湿度和水温的皮尔逊系数分别为0.27、-0.34和0.41。将3种冠层特征作为输入向量,采用逻辑回归算法判断水稻受褐飞虱侵害状况的测试集精准率为87.15%,召回率为86.54%,F1综合指标为86.55%。本文提出将气象因子与水稻的冠层特征数相结合,对水稻受褐飞虱侵害的冠层温度特征进行评估,可为水稻虫害的监测与诊断提供参考。
The change of canopy statistical temperature eigenvalue is one of the important index for crop pest identification.However,with the effects of environmental temperature and humidity fluctuations,when canopy temperature is used directly in the time series for pest evaluation,the healthy plants must be set for comparison.Therefore,the method is not operable in practical production applications.In order to find an effective method for evaluating the canopy statistical temperature eigenvalues of rice plants after brown planthopper infestation,the brown planthopper susceptible rice variety“TN1”was taken as the object,and two treatments of brown planthopper infestation and non-infestation were set.The infrared canopy was used to obtain the canopy of rice.The temperature eigenvalues were evaluated by using a machine learning classifier to evaluate the temperature characteristics of rice canopy-induced thermal images of rice canopy.In data analysis,three canopy statistical temperature eigenvalues extracted from the thermal images were used,and the features that best reflected the differences were selected.The cumulative difference of the canopy temperature coefficient of variation was 30.78.And then,combined with air temperature,relative humidity and water temperature,the logistic regression and support vector machine were used to fit the evaluation model.For determining brown rice planthopper damage by the logistic regression algorithm,when three canopy statistical temperature eigenvalues were used as input vector,the accuracy of the logistic regression test set was 87.15%,the recall rate was 86.54%,and the F1-measure was 86.55%.Support vector machine algorithm test set accuracy rate was 86.74%,recall rate was 86.90%,and F1-measure was 86.53%.In practical applications,the statistical eigenvalue of the canopy thermal image of rice can be obtained by calculating the air temperature,relative humidity and water temperature information to evaluate whether the inversion showed the invasion of brown rice planthopper.It was of great significance for the health monitoring and diagnosis of rice.
作者
刘又夫
肖德琴
刘亚兰
钟伯平
周志艳
LIU Youfu;XIAO Deqin;LIU Yalan;ZHONG Boping;ZHOU Zhiyan(Guangdong Engineering Research Center for Agricultural Aviation Application(ERCAAA),South China Agricultural University,Guangzhou 510642,China;National Center for International Collaboration Research on Precision Agricultural Aviation Pesticides Spraying Technology(NPAAC),Guangzhou 510642,China;College of Mathematics and Informatics,South China Agricultural University,Guangzhou 510642,China;College of Natural Resources and Environment,South China Agricultural University,Guangzhou 510642,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2020年第5期165-172,共8页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家重点研发计划项目(2018YFD0200301)
国家自然科学基金项目(31371539)
广东省重点领域研发计划项目(2019B020217003)。
关键词
水稻
褐飞虱
热红外技术
机器学习
统计学特征
气象因子
rice
brown rice planthopper
thermal infrared technology
machine learning
statistical eigenvalue
meteorological factors
