Relationship Between Spraying Droplet Size By Two Kinds Of Rotor UAV With Droplet Deposition Result

The rapid acquisition and processing of agricultural information is the prerequisite and basis for carrying out precision agriculture.Multispectral images can effectively reflect agricultural conditions such as crop leaf area index,diseases,insects and weeds due to their different wavebands.In recent years,precision agriculture,especially agricultural It has been researched and applied in areas such as remote sensing monitoring and precision agriculture and aviation.Low cost,fast acquisition,and convenient splicing processing are the basic requirements for the subsequent analysis of agricultural conditions based on multispectral images,and the conditions under which multispectral image acquisition systems can be widely used.In recent years,despite the rapid development of multi-spectral image acquisition systems based on drones,there are still inconveniences for users to change the spectral band in the application of existing systems,and it is not possible to plan routes according to user requirements according to the actual conditions of the field,and to cooperate with plant protection.The inconvenience of drone variable pesticide application and other problems,and the collected images also lack reliable,efficient and low-cost stitching software.In view of the above problems,this paper intends to carry out research on the acquisition control and splicing of UAV multispectral remote sensing images.The main research contents of this article are as follows:(1)Build a hardware platform for UAV multi-spectral remote sensing image acquisition and control.DJI Jingwei M100 UAV platform is equipped with embedded development board NVIDIA Jetson TX2,multi-spectral camera ADC Lite,digital transmission module,power module,and image format conversion.device.(2)Based on the Qt programming platform in the Linux environment,the ground station software uses the Google Map API to implement offline map development,and the flight control software uses the DJI Onboard SDK to implement functions such as UAV activation control,take-off and landing return,three-axis flight,flight missions,The communication between the ground station software and the flight control software is realized through a custom data transmission protocol.(3)A simple and reliable route planning algorithm is proposed: set the ground resolution and route overlap to calculate the drone altitude and route interval,and complete the route planning for the selected map area.At the same time,according to the definition of imaging parameters,set the ground resolution and heading overlap degree to calculate the photo interval.(4)Research on multi-spectral image stitching,including image preprocessing,image registration and image fusion,etc.This article focuses on the application research of image preprocessing and image registration methods.In the image fusion part,choose the weighted average fusion method.In the image preprocessing part,the HSL formula is used to calculate the brightness values of all the images,and the brightness values are divided into 5 units.The image corresponding to the middle brightness value is taken as the target matching image in the most occupied area,and then the histogram matching is used.Method for radiation correction.(5)In the image registration part,on the one hand,a method of image registration based on coordinate information is proposed.This method converts the calculation of the geographic coordinate offsets of two images into image coordinate offsets for image registration.An image registration test was performed based on coordinate information.The results showed that the stitching of the overlapping areas of the image was obviously misaligned.Then the results were tested based on gray value matching.The stitching results showed that the accuracy of the image stitching was greatly improved,but the stitching of the image was still There is a slight misalignment;on the other hand,a registration test based on image features is carried out to select the number of extracted feature points,feature extraction time,effective matching rate,and correct matching rate to measure the performance of the feature point detection algorithm.The results show that the effective matching rate of the SURF algorithm The average value of the correct matching rate is 82%and 32% respectively,and the average number of feature points and feature detection time are 3194 and 2.63 s respectively.Since the purpose of this study is to perform fast and high-accuracy stitching,comprehensive consideration is given to the selection of SURF algorithm combined with KNN coarse matching and RANSAC fine matching to complete image registration.Finally,based on the above research results,complete the design of the ground station software and flight control software.Through multi-spectral aerial flight test,the stitching of the flight trajectory map and the collected 98 multi-spectral remote sensing images shows that the actual flight trajectory of the UAV can fly along the planned route,and the resulting orthophoto images are ideally integrated,which proves that the system has a high The feasibility and stability of the system can achieve the expected results,and can provide multi-spectral analysis images of better quality for precision agriculture.