Research On Key Technologies Of Greenhouse Mobile Robot Mapping And Positioning Based On Visual SLAM

The greenhouse can artificially create an environment suitable for plant growth to meet people's demand for agricultural products in different seasons.At present,my country's greenhouses are still dominated by manual operations.Farmers have high physical labor intensity,high repetition rate,and harsh working environment,which cannot meet the production needs of modern greenhouses.The greenhouse robot can solve the above-mentioned problems to a great extent.The core problem for the greenhouse robot to move autonomously is positioning and mapping.Therefore,the research on the positioning and mapping of greenhouse mobile robots and the improvement of positioning and mapping accuracy are of great significance to the realization of unmanned modern greenhouses.This paper focuses on the impact of uneven lighting in greenhouses on image quality,the fusion of visual SLAM and IMU sensors,and the goal of improving the positioning and mapping accuracy of greenhouse mobile robots.Greenhouse mobile robot positioning and mapping provides solutions and core supporting technologies.The specific research content is as follows:First,in order to reduce the influence of the distortion error and zero bias error of the camera and the IMU sensor on the system,the camera and the IMU are calibrated.Use Kalibr?Allen toolkit and imu?utils toolkit to calibrate the camera and IMU respectively,and then calibrate them jointly to obtain the sensor's own error and improve the system accuracy.Secondly,in view of the problem that uneven illumination in the greenhouse affects the extraction and matching of image feature points,a greenhouse feature point extraction algorithm based on image enhancement is proposed to preprocess the images collected by the camera.Experiments have proved that this method can effectively improve the influence of uneven illumination on the feature point extraction,and increase the number of image feature points extraction and matching.Then,in view of the problem of mobile robot positioning failure caused by pure visual SLAM in the greenhouse,such as being easily blocked and rotating too fast,visual-inertial fusion SLAM is used to improve the positioning accuracy of mobile robots and improve the robustness of the system.The camera and IMU adopt a tightly coupled method for data fusion.The sliding window algorithm is used in the back end to reduce the amount of calculation of the system.Experimental verification results show that visual-inertial fusion SLAM can effectively improve the positioning accuracy of greenhouse mobile robots,and provide a basis for subsequent mapping.Finally,research on the establishment of a navigation map for the greenhouse mobile robot.Using point cloud splicing will build a dense point cloud map.Then the dense point cloud map is filtered in an offline manner to remove outliers and redundant points.Finally,use the function package in the ROS system to convert the dense point cloud map to an occupancy octree map.The experimental verification results show that the greenhouse occupancy octree map can be successfully constructed,which lays the foundation for subsequent navigation.