Research On Feature Extraction And Classification Of Street Trees Based On UAV Incline Photogrammetry

Street trees play an important role in urban greening and garden greening.Street trees are an important link in the construction of urban gardens,and have an important impact on the city's appearance and appearance.Therefore,the choice of street tree species is the top priority in urban gardens.The traditional sidewalk tree species survey adopts the method of manual field investigation,which is time-consuming and labor-intensive,and the survey efficiency is low,and it is difficult to meet the requirements of the sidewalk tree species survey.As a newer technology,ground-based lidar scanning technology has been widely used in tree species classification.This technology can quickly and efficiently obtain various forest parameters,and then realize the identification of tree species.However,the groundbased lidar scanning technology has instruments that are expensive,cumbersome,and inconvenient to carry.In actual operation,the operation of moving the station multiple times is required,resulting in increased labor costs.With the development of technology,especially the advancement of drone technology and photogrammetry technology,the tilt-based photogrammetry technology based on drone is becoming more and more widely used in agriculture,industry and construction industry.The technology was introduced into the investigation of forest resources,and research was conducted on how to use drones to extract forest parameters and how to quickly conduct forest resource investigations.However,the application of UAV tilt photogrammetry technology in practice is still mainly focused on the mapping of urban buildings,and its application in forestry is still relatively limited,especially in the field of street tree species classification.Therefore,it is necessary to study the feasibility of the UAV tilt photogrammetry technology in the classification of street tree species,and to explore the technical route and operation process to realize the technology.This article is based on the UAV tilt photogrammetry technology and uses the DJI Phantom4 RTK drone to obtain the tilt road photography photos of the target roadway trees.The network RTK function carried by this type of UAV can omit the work of setting control points in a small-scale measurement,which can reduce the workload of the field and ensure the positioning accuracy of the obtained data.Use the Pix4 Dmapper software to process the obtained oblique photogrammetry photos to generate a three-dimensional point cloud of the target road,use manual manual extraction to obtain the point cloud of a single street tree,and output the three-dimensional point cloud data in XYZ format.Using code written based on the Python language to extract the single tree measurement factors from the three-dimensional point cloud data.Compared with the actual measurement,the tree height extraction accuracy is 90.82%,the east-west crown amplitude extraction accuracy is 83.46%,and the north-south crown amplitude extraction accuracy is 83.89%.Based on the extracted single-tree tree-measuring factors,a machine learning method is used to conduct a classification study on street tree species.It is found that in this study,the four methods of KNN,random forest,support vector machine,and BP neural network have a higher accuracy rate for tree species classification,which are 84%,82%,78%,and 76%,respectively.Identification of street tree species.