Estimation of Landscape Element Distribution using Semantic Segmentation and DBSCAN

Abstract

In this research, we propose a new measurement method that combines deep learning technology and statistical processing for the purpose of analyzing the uneven distribution of landscape elements in urban areas. Specifically, in order to uniformly acquire landscape elements as three-dimensional information, we first present a method that utilizes multiple image recognition technologies. Then, the street-tree distributions in different urban areas such as downtown and new town are extracted and their characteristics are compared. The proposed method has the advantages of both actual observation and quantitative grasping, and has the potential to analyze various landscape elements in more detail.