Intelligent Moisture Prediction Modeland Irrigation Decision System Based On UAV Remote Sensing

There is a serious shortage of freshwater resources in China.Compared with the irrigation technology and the irrigation management in developed countries,they are still relatively backward which further changes required in the shortage of freshwater resources.As the continuous expansion of the arable land and the emergence of medium and large-sized agricultural irrigation areas.The requirements for the precise irrigation technology of farmland is much higher.The unmanned aerial vehicle(UAV)remote sensing technology has achieved the whole coverage information collection from point to region,which can provide the powerful data guarantee for the precise irrigation of the large-scale farmland irrigation.Therefore,it is necessary to establish an intelligent prediction model of soil moisture and develop an automatic processing system for the infrared thermal images and a decision-making system for the crop water demand.It is of great significance to improve and increasing tension of freshwater resources.The main research work and achievements of this thesis are as follows:(1)The correlation between crop canopy temperature,related meteorological parameters and soil water content inside the farmland was studied,and an intelligent prediction model of soil water content was established.Collected infrared thermal imaging of tea gardens,field weather station and target the soil moisture data simultaneously.Image processing algorithms such as Canny edge detection,adaptive median filtering,and mean filtering were used to eliminate soil background and noise pollution.Extracted canopy temperature data from the processed image.The PCA method was used to reduce the dimension of the collected data set,and the RBF prediction model of soil moisture was established by using the data sets before and after the dimension reduction.The study shows that it can effectively improve the accuracy of the target canopy temperature data by removing the noise and soil background in the image.Comparing the accuracy of these two soil moisture prediction models(RBF and PCA-RBF),the results show that the accuracy of these two prediction models are almost unchanged.(2)The related technologies of infrared thermal imaging stitching,background removal,and temperature extraction were investigated while a new automatic processing method for the remote sensing infrared thermal image was presented.Summarized the traditional SURF stitching method and made improvements to realize the stitching of multiple pictures.Converted image stitching result into grayscale images.Removed the soil background by identifying the gray level difference between the soil background and the crop canopy.Exploring the functional relationship between image gray data and temperature data.Used the gray data to extract the target canopy temperature data.Finally,the image processing methods of stitching,background removal and temperature extraction were integrated to establish an infrared thermal imaging processing system.The study results show that the stitched pictures can be efficiently obtained through this system and the stitched pictures are without gaps.The soil background can be completely eliminated.The reliability of the obtained canopy temperature data can reach more than 99%.This system can provide accurate canopy temperature data for subsequent irrigation decisions quickly.(3)The physical meaning of the theoretical model of crop water stress index(CWSI)and soil water correction coefficient(Ks)were studied.Established a new decision-making model of farmland irrigation quantity based on CWSI and Ks,and achieved automated decision-making.Finally,the system was experimentally verified.From the perspective of the decision-making process,the decision-making method has a strong theoretical basis and practicability.It can directly estimate the water demand of farmland.Judging from the decision results,the correlation coefficient between CWSI and crop stomatal conductance(Gs)is above 0.75.The correlation coefficient between water demand per unit area and crop Gs is above 0.8.The decision result has high prediction accuracy,which can be used for subsequent irrigation as reliable data support.(4)Researched and built a field irrigation monitoring system based on 4G network.Starting from the design of each subsystem of the irrigation monitoring system and the overall system architecture,discussed the subsystems and the whole system comprehensively.Completed the writing of the code of irrigation execution,data acquisition,etc.debugged each subsystem,and conducted on-site overall test.The test results show that the communication success rate of the 4G module is 100%,and the communication delay is acceptable(maximum 100ms).The host computer can100% realize control of the solenoid valve and collect flow data.The overall operation of the system is stable.It can meet the actual requirements.The operators can operate the system and access relevant irrigation data anytime and anywhere.