Construction of 3D Digital Reconstruction System of Urban Landscape Spatial Pattern Based on Deep Learning

The three-dimensional digital reconstruction technology can obtain the three-dimensional renderings of the spatial pattern of urban garden landscapes, which increases the scientificity and timeliness of garden design. The visual perception of urban landscape is the basis of spatial satisfaction and the quality of residents' travel, and the research on environmental perception oriented by visual resources is an important content in landscape planning. The 3D landscape index was initially proposed and constructed in the study of urban 3D landscape, but it was not applied in practice. An effective three-dimensional digital reconstruction system of urban garden landscape spatial pattern is of great significance for improving the design level of urban garden landscape spatial pattern. However, the use of high-resolution remote sensing images to extract building 3D information is a huge workload, and has been developed in recent years. In the past, the reconstructed urban landscape lacked authenticity. Some scholars used laser scanner to carry out three-dimensional digital reconstruction with high modeling accuracy. However, the reconstructed scene does not have texture information, has poor authenticity and costs. The human-oriented perspective quantifies and identifies the existing greenway landscape features and gives optimization suggestions, which is a direction to be developed in the greenway transformation and promotion in China. This paper mainly aims at in-depth learning to extract the spatial pattern features of urban landscape images due to the lack of feature matching algorithm in the three-dimensional digital reconstruction system.