Extraction Of Rice Canopy Temperature Characteristics Based On Thermal Image And Evaluation Method Of Brown Planthopper Stress

Brown Rice Planthopper(BRPH)is one of the main migratory pests of rice.It is usually harmed in the middle and lower parts of rice.It has the characteristics of long distance of migration and concealment of harmful parts,which seriously threatens the production of rice.This topic intends to use thermal image method to obtain the canopy temperature characteristics of rice,and evaluate the damage of brown planthopper,which can provide a basis for early monitoring and early warning of BRPH damage,and has certain practical application value for improving the timeliness of BRPH prevention and control..Based on the analysis of domestic and foreign insect evaluation and prediction methods,based on the mechanism and characteristics of BRPH damage to rice,thermal infrared imaging technology was used to study the changes of rice canopy temperature and its relationship with meteorological factors after BRPH invasion.Based on the evaluation method of BRPH invasion based on canopy thermal image and meteorological factors,the differences in the time series of canopy temperature characteristics between different resistant rice varieties before and after the damage of BRPH were compared.1.The segmentation method of rice canopy thermal infrared image under complex background conditions was studied.The kmeans algorithm was used to segment the canopy thermal infrared image.The kappa coefficient,confusion matrix and rice canopy coverage were used to evaluate the segmentation effect of the kmeans algorithm.At the same time,compared with the otsu algorithm,the results showed that the kmeans algorithm segmentation average The kappa coefficient is 78.92%,which is higher than the otsu algorithm.The average error is 28.42%,which is lower than the otsu algorithm.The kmeans algorithm is better than the otsu algorithm.On the basis of image segmentation,the eigenvalues such as mean,median,maximum and mode of the canopy temperature are extracted.Compared with the canopy temperature feature extracted by manual segmentation,the kmeans algorithm and the otsu algorithm segmentation extract the crown.The difference between the two characteristics of the mode temperature and the maximum value of the layer temperature is not significant(at the 0.05 significance level,sig=0.0089 and sig=0.016),and the maximum average error of the kmeans algorithm and the otsu algorithm is 0.258 °C,the mode average The errors are 0.138 °C and 0.15 °C respectively;the mean and median values of canopy temperature extracted by kmeans algorithm and otsu algorithm are significantly different(at 0.05 significant level,sig=0.0089 and sig=0.0161),kmeans The mean and median errors of the algorithm extraction were 0.261 ° C and 0.240 ° C,both significantly lower than the otsu algorithm.2.A method for evaluating the BRPH invasion based on the fusion of canopy thermal images and meteorological factors was determined.Since the absolute temperature of rice canopy is greatly affected by meteorological conditions,it is difficult to estimate the invasion of BRPH by absolute temperature in time series.In order to evaluate whether the change of canopy temperature in rice is caused by BRPH invasion,thermal infrared is used.The imaging technique compares the canopy temperature of the BRPH invading rice and the uninvaded rice,and uses the temperature and humidity sensor to record the atmospheric temperature,humidity and reference water temperature at the moment of collecting the thermal infrared image,and fuses the canopy thermal image with the meteorological factor.Reconstructed standard deviation of canopy temperature(CTSD),canopy temperature range(CTR),maximum temperature of canopy temperature(CTM),and canopy canopy temperature coefficient of variation CTCV(CTCV)evaluation index,through the establishment of feature engineering to select input,using k-near neighbor classification algorithm(KNN,k-Nearest Neighbor),logistic regression(LR,Logistic Regression)and support vector machine(SVM,Support The Vector Machine method is used to evaluate the invasion of the BRPH.The results showed that on the 10 th and 14 th day of the BRPH damage,the four evaluation indexes between the damaged and unaffected rice canopy reached significant difference(at 0.01 significant level),and the F of the CTCV index was The two days are the biggest.The F between the two between 9:30 and 15:00 is 168,which is much larger than the F of other time periods.Correlation analysis of air temperature,relative humidity and reference water temperature on CTCV showed that the three had a great influence on the degree of CTCV difference,and the correlation coefficient of reference water temperature was the highest,reaching 0.854,indicating the extraction of rice canopy thermal image.The CTCV indicator changes with changes in ambient temperature and relative humidity.When monitoring the BRPH invasion in the actual field environment,it is often difficult to set up the unaffected sample as a control.In order to establish the BRPH invasion monitoring model in the field environment,CTCV,leaf temperature,air temperature,humidity and water temperature are used as inputs.Based on the feature quantity,a monitoring model of BRPH invasion based on support vector machine SVM is established.The model verification test has obtained 242 samples.The training set and test set of the model are divided into 7:3.The test results show that the training The set recognition accuracy rate is 82.84%,and the test set correct rate is 80.82%.3.The differences in time series between the comprehensive evaluation indexes of canopy temperature of different resistant rice varieties before and after the damage of BRPH were compared.Thermal infrared imaging was used to monitor the canopy temperature of different insect-resistant rice cultivars under the control of BRPH,and the time series was monitored under controlled conditions.The insect-resistant rice varieties were treated with “excellent” and non-insect resistant rice varieties were adopted."TN1".The results showed that after 23 days of BRPH damage,the average absolute temperature of the canopy were increased to different degrees compared with the first day.The “excellent” increased by 0.676°C and the “TN1” increased by 0.680°C.The difference in temperature rise is not significant(at a significant level of 0.01);in order to find differences in thermal image thermal image characteristics between resistant rice varieties and susceptible rice varieties after exposure to BRPH,CTSD,CTR,CTCV And four comprehensive evaluation indicators such as CTM were analyzed.The results showed that the three comprehensive evaluation indexes of CTSD,CTR and CTCV of rice canopy thermal infrared images first increased and then decreased,and the above three The peak of the value is on the 13 th day of being harmed.CTM is the trend of decreasing first and then increasing,and the lowest value of CTM of both varieties is on the 13 th day of being harmed.On the 13 th day,the CTSD,CTR and CTCV of “TN1” were higher than “excellent” in the first 13 days of harm,and the CTM was lower than “excellent”;after 13 days “TN1” The CTSD,CTR and CTCV are lower than “excellent”,and the CTM is higher than “excellent”.The results of variance analysis were carried out on the results of each of the four comprehensive evaluation indicators of two different resistant rice varieties.The results showed that the four evaluation indicators did not reach significant difference(at 0.1 significant level),and among the four evaluation indicators.The F between C15 and 15 to 19 days is larger than the other three indicators.