Research On Reconstruction Technology Of Orchard Environment Topography And Landform

With the rapid development of agricultural modernization,there is an increasing demand for automated agricultural machinery,especially unmanned agricultural machinery,in the agricultural production field.Forest fruit production is an important area of ??agricultural production,and there is an increasing demand for unmanned farming machinery and unmanned picking machinery,and at the same time it is very sensitive to production costs.Unmanned technology mainly includes environmentaware technology,positioning and navigation technology and motion control technology,among which environment-aware technology is one of the basic technologies.At present,the technology of driverless vehicles is developing rapidly,but its environmental perception research is mainly aimed at the structured road environment and parking environment,and there is little research on the unstructured environment(if the garden environment).This paper aims at the low-cost requirements of agricultural production for unmanned agricultural machinery,and uses the low-cost and low-beam radar as the measurement tool to identify orchards,low vegetation and poles.Topographical reconstruction and map construction.The main research contents are as follows:According to the source of the noise point of the point cloud when the point cloud data is acquired,the straight-pass filtering method of the distance limitation is used to limit the processing range of the point cloud data,and the local window setting is realized.The Gaussian filtering method based on data statistics is used to remove outliers,and the voxel mesh filtering method is used to measure the distance of the point cloud to achieve the purpose of controlling the number of cloud points without changing the shape of the measured object.For the tilt of the image caused by the shaking of the measuring device,the tilting of the point cloud plane is implemented by vector fork multiplication.Aiming at the scarcity of information when measuring less beam laser lidar,this paper proposes a method based on the point cloud side projection analysis method to realize the recognition of fruit trees,low vegetation and utility poles in a local area,and complete the window object recognition.In the method,the point cloud data is projected on the YOZ coordinate plane according to the measurement angle of the radar;the eight-neighbor search algorithm is used to cluster the ground point projection data to realize the point cloud partitioning;the projection surface is meshed and The object point statistics in the grid are used to judge the empty grid and the nonempty grid;the low-to-high layer-by-layer scanning method is used to judge the nonempty grid adjacency relationship;according to the adjacent relationship,the adjacent non-empty grid is used.Elevation values,span values,and their mutations enable the identification of fruit trees,low vegetation,and utility poles within the window.The experiment proves that this method can effectively realize the identification of fruit trees,low vegetation and utility poles in the orchard environment.Aiming at the disorder and abruptness of point cloud data in orchard environment,an improved adaptive combination registration algorithm is proposed to register different point cloud data to realize global terrain and terrain perception.In this method,the NDT algorithm is used to initialize the adjacent window map according to the topographical features of the local environment.The ICP is used as the exact registration to increase the curvature as the matching element,and the adaptive point cloud matching algorithm is used to perform the precise registration of the adjacent window map.