Research On 3D Reconstruction Method Of Citrus Based On 3D Lidar

Citrus is an important economic crop in the south,and the three-dimensional orange tree model has important research significance for the realization of intelligent orchard machinery.At present,the data obtained by the 3D laser scanner has high accuracy,but it is expensive,has a long acquisition time,and is not mobile.The mobile 3D laser radar has a wide range,mobility,and an affordable price.It is obtained in crop phenotype measurement.Attention and application.The M8 type 3D laser radar of QUNERGY company was selected to scan and obtain the three-dimensional space information of orange tree and to reconstruct the three-dimensional model of orange tree based on the point cloud.Test to build an M8 point cloud data collection platform,complete mobile data collection,and combine the point clouds according to the type of point cloud filter processing;put forward the center point rotation combined with ICP algorithm splicing method;use the shortest path algorithm to connect the skeleton points,based on the skeleton Line to achieve branch reconstruction.The main research work and conclusions are as follows:(1)To acquire point cloud data,build an orange tree point cloud collection platform with M8 as the core.The acquisition platform is mainly composed of linear modules,servo motor systems,controllers,M8 and other hardware.The servo motor system accurately controls the movement of the lidar on the linear module,collects single-sided point cloud data of the orange tree point cloud,and merges the point cloud through the movement speed and the frame number information.According to the point cloud data type,determine the combined filtering method to remove the outliers,to obtain a higher precision point cloud,and meet the next point cloud stitching requirements.(2)Propose a method of matching the ICP algorithm by rotating the angle between the center points of the point cloud and collecting points.Due to the complexity of the tree point cloud,it is impossible to extract high-quality point cloud feature vectors;setting target collection has limitations on the acquisition of all-round data of the target,which may cause large model errors due to incomplete data.The angle is combined with the ICP algorithm matching method.Find the center point of the point cloud through the algorithm,and combine the angle between the collection points to carry out the spatial transformation of the point cloud to complete the point cloud rough stitching.The ICP algorithm is used to perform the secondary matching on the coarse splicing point cloud to improve the splicing accuracy of the point cloud.The relative error of the three-dimensional size of the model obtained through the experimental method is within 7.5%,and the experimental results meet the requirements of the three-dimensional point cloud to extract skeleton information.(3)Three-dimensional model reconstruction.Select the spatial thinning method to process the point cloud data,and retain the characteristics of the point cloud model to the greatest extent.The cluster point is selected by the clustering algorithm to represent the point cloud skeleton point,and the skeleton point is connected by the shortest path algorithm to obtain the skeleton line that basically matches the point cloud model.The experiment is based on the reconstruction of the skeleton line through the pipeline theoretical model and the generalized cylinder method.The thickness of the branch is mainly controlled by the pipeline theoretical model.The branch is represented by a cylinder,and the branches are rendered with leaves and textures to obtain fruit trees with good visual effects.model.