基于热图像特征对白粉病感染玫瑰叶进行的早期检测
|
摘要
通过使用数字红外热成像仪观察感染和未感染的玫瑰植物叶片表面温度的变化,发现感染区的叶片温度上升2.3℃。此外通过对健康和感染的叶片进行分类,选取最佳试验叶片并观测它们的热特征;利用绝对温度测量,选择其温度最大值、最小值、中位数、最大温差、标准偏差,拟合数据到标准正态分布和拉普拉斯分布曲线,然后,通过神经模糊分类器来识别感染和健康的叶子;最后利用k均值聚类方法获得原始参数和模糊规则,在分类器的8个簇进行训练和测试,准确率分别达到了92.55%和92.30%。试验结果表明,干旱对健康叶片有不利影响,在干旱条件下健康的叶片正面预测价值和特异性指数值相应降低,而在黑暗中叶片的性能没有显著影响。
The surface temperature of infected and uninfected rose plants was observed by digital infrared thermal imager.The leaf temperature of the infected area was increased by 2.3 ℃.In addition,by classifying healthy and infected leaves,select the best experimental leaves and observe their thermal characteristics;use absolute temperature measurement to select their temperature maximum,minimum,median,maximum temperature difference,standard deviation,and fit the data.To the standard normal distribution and the Laplacian distribution curve,then the neural fuzzy classifier was used to identify the infected and healthy leaves;finally,the k-means clustering method was used to obtain the original parameters and fuzzy rules,and the 8 clusters of the classifier were trained.And testing,the accuracy rate reached 92.55% and 92.30%.The results showed that drought had an adverse effect on healthy leaves.Under drought conditions,the positive predictive value and specificity index of healthy leaves decreased correspondingly,while the performance of leaves did not significantly affect in the dark.
The surface temperature of infected and uninfected rose plants was observed by digital infrared thermal imager.The leaf temperature of the infected area was increased by 2.3 ℃.In addition,by classifying healthy and infected leaves,select the best experimental leaves and observe their thermal characteristics;use absolute temperature measurement to select their temperature maximum,minimum,median,maximum temperature difference,standard deviation,and fit the data.To the standard normal distribution and the Laplacian distribution curve,then the neural fuzzy classifier was used to identify the infected and healthy leaves;finally,the k-means clustering method was used to obtain the original parameters and fuzzy rules,and the 8 clusters of the classifier were trained.And testing,the accuracy rate reached 92.55% and 92.30%.The results showed that drought had an adverse effect on healthy leaves.Under drought conditions,the positive predictive value and specificity index of healthy leaves decreased correspondingly,while the performance of leaves did not significantly affect in the dark.
引文
[1] AL-HIARY H,BANI-AHMAD S,REYALAT M,et al.AlRahamneh,fast and accurate detection and classification of plant diseases[J].Int J Comput Appl,2011(17):31-38.
[2] JANG J S R,SUN C T,MIZUTANI E.Neuro-fuzzy and soft computing[M].Prentice Hall,Upper Saddle River,NJ,1997.
[3] CHAERLE L,van CAENEGHEM W,MESSENS E,et al.Presymptomatic visualization of lant-virus interactions by thermography,Nat[J].Biotechnol,1999(17):813-816.
[4] LINDENTHAL M,STEINER U,DEHNE H W,et al.Effect of downy mildew development on transpiration of cucumber leave svisualized by digital infrared thermography[J].Phytopathology,2015,95:233-240.
[5] MOSHOU D,BRAVO C,OBERI R,et al.Plantdisease detection based on data fusion hyper-spectral and multi-spectral fluorescence imaging using Kohonen maps[J].Real-Time Imag,2016(11):75-83.
[6] BABU G S,SURESH S.Meta-cognitive RBF network and its projection based learning algorithm for classification problems[J].Appl Soft Comput,2013(13):654-666.
[7] AYRES P G.Patternsof stomatal behaviour,transpiration and CO2 exchange in pea following infection by powdery mildew (Erysphe pisi)[J].J Exp Bot,2016(27):354-363.
[8] CALDERON R,NAVAS-CORTES J A,ZARCO-TEJADA P J.Early detection andquantification of Verticillium wilt in olive using hyperspectral andthermal imagery over large areas[J].Remote Sens,2015(7):5584-5610.
[9] CETISLI B.The effect of linguistic hedges on feature selection:Part2[J].Expert Syst Appl,2010(37):6102-6108.
[10] BRAVO C,MOSHOU D,WEST J,et al.Early disease detection in wheat fields using spectral reflectance[J].Biosyst Eng,2003,84:137-145.
[11] CALDERON R,NAVAS-CORTES J A,ZARCO-TEJADA P J.Early detection andquantification of Verticillium wilt in olive using hyperspectral andthermal imagery over large areas[J].Remote Sens,2015(7):5584-5610.
[2] JANG J S R,SUN C T,MIZUTANI E.Neuro-fuzzy and soft computing[M].Prentice Hall,Upper Saddle River,NJ,1997.
[3] CHAERLE L,van CAENEGHEM W,MESSENS E,et al.Presymptomatic visualization of lant-virus interactions by thermography,Nat[J].Biotechnol,1999(17):813-816.
[4] LINDENTHAL M,STEINER U,DEHNE H W,et al.Effect of downy mildew development on transpiration of cucumber leave svisualized by digital infrared thermography[J].Phytopathology,2015,95:233-240.
[5] MOSHOU D,BRAVO C,OBERI R,et al.Plantdisease detection based on data fusion hyper-spectral and multi-spectral fluorescence imaging using Kohonen maps[J].Real-Time Imag,2016(11):75-83.
[6] BABU G S,SURESH S.Meta-cognitive RBF network and its projection based learning algorithm for classification problems[J].Appl Soft Comput,2013(13):654-666.
[7] AYRES P G.Patternsof stomatal behaviour,transpiration and CO2 exchange in pea following infection by powdery mildew (Erysphe pisi)[J].J Exp Bot,2016(27):354-363.
[8] CALDERON R,NAVAS-CORTES J A,ZARCO-TEJADA P J.Early detection andquantification of Verticillium wilt in olive using hyperspectral andthermal imagery over large areas[J].Remote Sens,2015(7):5584-5610.
[9] CETISLI B.The effect of linguistic hedges on feature selection:Part2[J].Expert Syst Appl,2010(37):6102-6108.
[10] BRAVO C,MOSHOU D,WEST J,et al.Early disease detection in wheat fields using spectral reflectance[J].Biosyst Eng,2003,84:137-145.
[11] CALDERON R,NAVAS-CORTES J A,ZARCO-TEJADA P J.Early detection andquantification of Verticillium wilt in olive using hyperspectral andthermal imagery over large areas[J].Remote Sens,2015(7):5584-5610.