SLAM Technology For UAV Application Under Canopy Scenario Based On Fusion Of Vision And Inertial Measurement Unit

The remote sensing method of UAV monitoring presents distinctive advantages such as real-time,fine and distant characteristics.This novel method plays an important role in modern forestry information monitoring field and it has been widely used in upper crown monitoring scenarios.There is abundant forest environment information in under-canopy environment.Therefore it is significant to monitor the information that exists under the canopy.Upper canopy environment is open to receive GNSS signals but for the monitoring under the canopy,the problem of GNSS signal loss is severe.This thesis aims to solve the problem that the drone can not conduct monitoring when the navigation under the canopy is challenging.Towards the requirement of carrying out practical flight mission under the canopy,this thesis proposes a method of sensor information fusion based on RGB-D camera and inertial measurement unit using SLAM technology,which realizes the map generation and drone positioning of the quadrotor UAV in the complex environment under the canopy.The main contribution accomplished in this thesis is listed as follows.(1)A mathematical model of a quadrotor UAV was established.According to the flight principle of the quadrotor UAV,the model parameters of the UAV were analyzed and the dynamics model of the quadrotor UAV was studied through Newton's Euler method.Furthermore,the system state equation of the quadrotor UAV was obtained and the design of the PID controller was carried out.(2)Research on the method of image feature point extraction and matching was studied.Firstly,the RGB-D camera was used to efficiently acquire the characteristics of the landmark information in the image to complete the image information acquisition process.Then the ORB feature points were used to extract and match the acquired image feature points.Finally,the RANSAC algorithm was improved in the feature matching back-end and the mismatching point screening was performed.Apart from the realization of real-time requirements,the accuracy of feature point matching was improved.(3)A solution of RGB-D Camera and IMU Fusion Location Algorithm was proposed herein.The IMU pre-integration theory and the tightly coupled visual inertial fusion method were used to estimate the position of the key frames extracted by the image.The nonlinear optimization method of the inter-sliding window method was used to fuse the IMU and the visual information.Finally,regarding the practical requirements of the flight monitoring mission in the environment under the canopy,a quadrotor UAV SLAM system platform based on RGB-D vision and IMU inertial information fusion was built and the positioning and map of the drone canopy was carried out.The experimental results of the platform showed that the proposed algorithm is more accurate and robust than the ORBSLAM algorithm,which can meet the requirements of UAV positioning and map construction under the canopy scenario.