Extracting Street Componments And Measuring Three-dimensional Visible Quality Of Urban Street Space Based On Mobile LiDAR Point Clouds

China's urbanization process has entered a new stage,emphasizing the development concept and principles of"people-oriented,optimized layout,ecological civilization,and cultural heritage."The quality of urban space is an important part of the evaluation of human settlements.It represents the physical suitability of the space environment for people and a spatial perception of the built environment of the city.It is also a visual measurement that is highly related to the characteristics of the urban physical environment.The improvement of urban space quality not only plays an important role in improving the urban ventilation environment,dispersing low-altitude pollutants,and reducing the heat island effect,but also has an important impact on the behavior,cognition and mental outlook of urban residents,which has become one of the focuses of current urban research.As an important urban public space closely related to people's lives,streets include not only the roads themselves,but also materials and facilities related to the streets,such as buildings on both sides of the street,street trees,service facilities,shops and open spaces.Streets are an important carrier that concentratedly reflects the quality characteristics of urban space.Therefore,assessing the quality of urban street space and its distribution characteristics is of significance for scientifically and practically improving the level of urban planning and construction and creating a new type of intensive,intelligent,green and low-carbon town.Due to the restriction of measurement methods and data sources,the existing research on street space quality lacks quantitative analysis of street three-dimensional visible material elements and relatively complete evaluation techniques,and there is a lack of street three-dimensional visible space quality evaluation indicators from a stereoscopic perspective.The vehicle-borne laser scanning system can quickly collect the surface information of urban objects such as roads and surrounding buildings,trees,vehicles,etc.due to the characteristics of its platform and operation method.It provides the possibility for the feature extraction of material elements in street space and the refined measurement and evaluation of street space three-dimensional visible quality.To this end,for the scientific problem of how to realize refined three-dimensional visual space quality measurement,the paper carried out research on street space three-dimensional visible quality based on mobile laser scanning data.The grid scale and street scale space three-dimensional visible quality measurement and evaluation are realized by using mobile lidar data and verified by two research areas in Ningbo.The main research contents and results of this article are as follows:(1)A feature classification method based on local point cloud features and Gradient Boosting is proposed,which can classify seven types of features(ground,buildings,trees,fences,vehicles,poles,shrubs)in the mobile laser point cloud.Based on the denoising of the point cloud,the ground point cloud was extracted first,and then the dimensional features,three-dimensional features,point feature histogram features,density features,elevation features,projection area features,intensity,and color features of the point cloud were extracted.After induction and calculation,a 52-dimensional point cloud feature vector was constructed,and six feature types were automatically classified using the Gradient Boosting algorithm.After verification,the overall accuracy of the classification in the two research areas reached 89.70%and90.95%respectively,indicating that the method proposed in this study has a high accuracy.This method lays the foundation for further analysis of the space arrangement and form of urban street elements.(2)A calculation method of 3D visibility of streets based on voxels in a point cloud environment is proposed to analyze the 3D visible space size and openness distribution at different locations in the street environment.It is of great significance to understand the openness of urban streets.Based on the characteristics of mobile LiDAR that can accurately detect the three-dimensional structure of streets,this study proposed a three-dimensional visibility measurement method based on voxels in a complex point cloud environment of large-scale urban streets,as well as a quantitative method to achieve three-dimensional openness of the volume index,which are implemented to accomplish the Visual skyline and 3D visual volume representation.The calculation method of"voxel-ray-polyhedron"is put forward.Based on the fast intersection algorithm of sight line and voxel and the volume calculation method based on the grid model,the calculation of 3D visibility is realized.The results show that,compared with the traditional two-dimensional visual analysis index and three-dimensional visual volume calculation method,the three-dimensional visual space calculation method and volume index proposed in this study can better reflect the distribution of the street space three-dimensional openness,which provides an important technical basis and method support for the measurement and evaluation of street space quality.(3)A set of evaluation indicator system of street three-dimensional visible space quality using point cloud and visible space quality measurement process and method based on machine learning are proposed.The street visible space quality elements and their distribution are analyzed,and the street space quality is classified.First,this paper selects three-dimensional visible space quality evaluation indexes based on street material elements,including seven indicators:volume index,street green index,sky index,enclosure index,road motorization index,car occurrence index and diversity;Then an ELO scoring algorithm is used by experts to compare the selected377 sample points,and the quantitative score of sample photos is calculated;Finally,the random forest machine learning method is used to automatically score the grid scale three-dimensional visible quality.By comparing with the accuracy of the actual score of experts,it is found that the accuracy of the proposed method is R~2=0.86,indicating that the method has high practicability and achieved refined quality measurement of streets in two research areas of Ningbo.The results show that areas with high visual space quality scores tend to have larger sky degree,street green visibility rate,road motorization degree and car occurrence rate,while areas with low visual space quality scores tend to have larger volume index,interface closure and diversity.Different from previous studies,this paper not only quantifies the subjective evaluation indicators of street space quality from the perspective of three-dimensional visual material elements,but also introduces the machine learning method to complete the large-scale refined space visible quality measurement more quickly and objectively.The proposed method can provide decision support for urban builders,and promote the construction of new urbanization and urban ecological civilization.