Objective Visual characteristics of landscape space are one of the criteria for measuring the quality of landscape space. The in-depth investigation of the quantitative analysis method for visual characteristics of landscape space will assist designers in accurately recognizing the spatial form of landscape space, and provide support for the design of high-quality landscape space. Previous scholars have conducted numerous studies on the association between landscape visual characteristics and human perception, proposed different assessment indicators for describing visual characteristics of landscapes, and confirmed that the visual characteristics of landscape space play an important role in predicting people’s preference for a specific landscape space as well as their potential activity behaviors. However, the previous analysis methods based on 3D models or 2D photographs can hardly take into account the quantitative analysis of the spatial information of visual interface in both the plane and depth dimensions, which is less applicable in the research and design practice regarding the spatial visual characteristics of large-scale urban parks. Although the point cloud model is highly accurate, the quantification method based on point cloud voxelization entails long-time computing, and there are still some difficulties in the quantitative analysis of spatial visual characteristics of large-scale urban green landscapes. Therefore, this research aims to propose a quantitative analysis method for the visual characteristics of landscape space based on point cloud and visualization algorithms. On the basis of retaining the high-precision spatial data of the point cloud model, this method has high computational efficiency and can meet the demand for quantitative analysis of 2D and 3D visual characteristics.

Methods Based on previous research on the quantification of landscape spatial morphology, this research combines the point cloud visualization with the spatial analysis function in ArcGIS and calculates multiple visual indicators by simulating the visual interface of multiple viewpoints in landscape space, thus realizing the computation and visualization of the visual characteristics of landscape space. The method is mainly divided into four steps, including point cloud model data acquisition and processing, spatial unit delineation and viewpoint generation, visual simulation and characteristic indicator calculation, and visualization and analysis of the visual characteristics of spatial units. First, point cloud model data are acquired by virtue of unmanned aerial vehicle (UAV) oblique photography and classified into a total of seven categories: Trees and shrubs, ground cover, lawn, hard ground, water surface, buildings, and others. Second, spatial units are identified by slope analysis in ArcGIS, and viewpoints are generated using the fishing net tool created in ArcGIS. Using the point cloud visualization and depth image tools in Point Cloud Library (PCL), the visual interface of each viewpoint is simulated, and the depth value and semantic value corresponding to each pixel point of the 2D image are extracted, which can be computed for multiple visual characteristic indicators. Five visual characteristic indicators (extension, openness, complexity, green visibility, and sky ratio) mostly used in previous research, are selected as examples. Relevant calculation formulas and methods are developed according to the proposed method. Finally, upon calculation of the visual characteristic indicators of each viewpoint, corresponding viewpoint properties are added to the viewpoint object in ArcGIS, and the calculated indicator values are imported into the characteristic table. Additionally, the Kriging interpolation tool in ArcGIS is adopted to generate a color raster map of each indicator.

Results In order to verify the feasibility of the proposed method, Qinglvyuan Park in Najing is taken as an example to explore the application of the method. Four spatial units with an area larger than 0.5 hm² are extracted for quantitative analysis. From the results, the visual characteristics of the four spatial units differ significantly. The spatial distribution characteristics of each visual indicator in the spatial units are weakly correlated with the plane geometry of such units. Although the value of each visual characteristic indicator continuously changes in space, the distribution locations of the peak and valley values are related to the elements inside the space (e.g., internal trees). Compared with previous analysis methods, the proposed method can more comprehensively and intuitively show the distribution of the indicator values and continuous changes of multiple visual characteristics in space, and has the potential to comprehensively analyze multiple characteristics based on the quantitative analysis of multiple visual characteristics.

Conclusion The point cloud visualization method can replace the manual photographs taken in the landscape environment, and it can obtain spatial information on the depth of visual interface that cannot be fully covered by the photographs. Although it has certain advantages over traditional methods in terms of the efficiency and accuracy of visual interface simulation and analysis, there are some shortcomings in terms of the data acquisition method of the point cloud model, the rules of parameter setting, and the applicability to landscape spaces with different scales. This method has certain application prospects in the fields of exploring the correlation mechanism between the visual characteristics of landscape space and the behavioral preferences of crowds, predicting the behavioral activities of crowds, and designing the interaction between people and the environment.