Lodging Characteristics Of Rice Based On UAV Multi-source Information

Rice is one of the three major food crops in the world and an important food crop with a wide area of cultivation in China.Food security will be affected to some extent with the fluctuation of rice yield,so the rice production capacity is of great significance in terms of food security.Lodging is one of the most important factors affecting the high and stable yield of rice.Lodging not only affects the harvest of rice,but also results in the decrease of yield and quality.In recent years,with the innovation and development of technology,more and more high-quality and high-precision sensors have been developed,which also promote the rapid development of image information acquisition based on UAV to monitor rice lodging.In this study,a UAV equipped with a visible light camera was used to obtain RGB images of rice and process the images.The visible light atmospheric impedance color index(VARI),super green color index(ExG),super red color index(ExR),and vegetation were selected.Color index(CIVE),improved green-red color index(MGRVI)and red-green color index(RGRI)are used as research reference indexes.The thermal infrared imager carried by the drone is used to obtain thermal infrared images of rice and process and analyze the images.At the same time,this study also used the UAV equipped with a hyperspectral camera to acquire images of rice in lodging and non-lodging areas.Normalized vegetation index(NDVI),ratio vegetation index(RVI),enhanced vegetation index(EVI),structural insensitivity pigment index(SIPI)and atmospheric impedance vegetation index(ARVI)were selected as the reference for hyperspectral monitoring of rice lodging index.Finally,a training support vector machine(SVM)classifier is used to construct an accuracy evaluation model of the rice lodging monitoring method.The results show that MGRVI and ExG can be selected as appropriate color feature parameters for information reference and numerical extraction when using the RGB images to identify rice lodging areas and non-lodging areas.When monitoring the degree of lodging,ExG can be preferred as an index for information extraction.However,this method is greatly affected by nitrogen fertilizer,and the effect of low nitrogen fertilizer on color characteristic value is greater than that of high nitrogen fertilizer.When using thermal infrared to monitor the lodging of rice,the temperature of rice maintains a small variation range throughout the day.The temperature of the stem is usually higher than the temperature of the leaves.After lodging,the temperature of the canopy rises,and it rises by 5℃ in the summer with sunlight.The thermal infrared instrument equipped with a drone can successfully distinguish between lodging rice and non-lodging rice,with the highest accuracy rate being from 10 am to 4 pm.Therefore,thermal infrared image is an effective method for monitoring rice lodging,and it is less affected by rice varieties and nitrogen fertilizer treatment.You can choose 19℃ as the highest temperature characteristic value,when it is greater than 19℃,it is lodging rice,and if it is less than 19℃,it is normal rice;choose 17℃ as the lowest temperature characteristic value,when it is greater than 17℃,it is lodging rice,and if it is less than 17℃,it is Normal rice;choose 18℃ as the average temperature characteristic value,when it is greater than 18℃,it is lodging rice,and if it is less than 18℃,it is normal rice.After lodging,the composition and light transmittance of the canopy and the plant’s own physiological characteristics change,which affects the hyperspectral reflectance.Among the five vegetation indexes of NDVI,RVI,EVI,SIPI and ARVI,RVI can be selected as the best vegetation index for evaluating lodging rice.In RVI,1.90 can be selected as the characteristic value,which is a good distinction between lodging and non-lodging rice.When the value is greater than 1.90,it is non-lodging rice,and when the value is less than 1.90,it is lodging rice.At the same time,RVI is very significantly correlated with the degree of rice lodging.When the lodging level is higher,the value is lower,which has a good ability to characterize the degree of rice lodging.By analyzing the accuracy of single feature monitoring lodging,single image source lodging monitoring accuracy and multi-source information fusion lodging monitoring accuracy,it is found that the missed detection rate and false detection rate of rice lodging monitoring under the integration of multi-source information technology are obvious below single feature and single image source.