Google Street View와 딥러닝을 활용한 서울시 녹지 형평성 분석 : NDVI와 가로이미지 기반 녹지 산출방법과의 비교를 중심으로

기동환(Ki, Donghwan),김선재(Kim, Sunjae),이수기(Lee, Sugie) 대한국토·도시계획학회 2021 國土計劃 Vol.56 No.4

Urban green has various benefits, including promoting physical activity, improving residents’ health, and mitigating urban heat islands. Hence, urban green is considered essential for urban residents, but green inequity issues are being raised. Although several studies have analyzed green equity with the traditional measurement method, the conventional approach is limited in its inability to reflect the actual degree of the green exposure of residents. To fill this gap, this study aims to identify the actual green equity using the Green View Index (GVI), which can represent actual green exposure. This study utilized Google Street View (GSV) and computer vision techniques to measure the GVI. The normalized difference vegetation index (NDVI) and geographic information system (GIS) based green area variables, which are traditional green area variables, were used to compare these distributions with GVI. Furthermore, this study identified the degree of green equity through the relationship between the distribution of green variables and the vulnerable groups. In terms of statistical model, the spatial lag and spatial error models were used to control the spatial autocorrelation. The results of this study are as follows. First, there were significant distributional differences between traditional green variables and GVI. Specifically, traditional green variables were high in the periphery of Seoul. GVI, however, was shown as cold-spots in these areas and highly concentrated in Gangnam, Seocho, and Songpa-gu. Second, the GVI model showed a lack of street greenery where numerous vulnerable people live, unlike traditional green variable models. Specifically, low-income people tend to live in neighborhoods with less street vegetation. Therefore, the government should implement green supply policies for these target neighborhoods. Regarding the methodological perspective, the results indicate that the degree of green inequality may vary depending on the green measurement methods. Moreover, plans for the supply of green should be based on GVI that can represent the actual degree of the exposure of residents.

스트리트뷰를 이용한 녹색환경과 사회 경제 변수의 인지 능력에 대한 상관관계 연구 (은평구 코호트를 대상으로)

주유형(Joo, Yoohyung),박상윤(Park, Sangyoon),정재영(Jung, Jaeyoung),홍지완(Hong, Jiwan),고주연(Ko, Juyeon),조재림(Cho, Jaelim),김창수(Kim, Changsoo),허준(Heo, Joon) 한국측량학회 2023 한국측량학회지 Vol.41 No.6

Using Street-view this study measured the green index within a 500-meter radius of the residences of a cohort visiting Severance Hospital from Eunpyeung-gu. We examined the association of the green index and other individual factors(gender, age, income, and education) with cognitive ability. We found that neither the green index obtained from Street-view nor the EVI from satellite imagery showed a significant association with cognitive health. Instead, age, gender, and education level were revealed affecting cognitive health. As age decreased, being female, and having a higher education level were associated with better cognitive health. This suggests that individual factors are more closely related to cognitive ability than neighborhood factors such as the green index. However, to assess cognitive ability more objectively, alternative measurement methods such as brain imaging might be necessary. Additionally, this study solely focused on Eunpyeong-gu which exhibited a small variation in street view green indices. Further research should consider expanding the study area and adding various independent variables.

녹시율과 정규화 식생지수에 이용한 도시 가로녹지공간에 관한 연구 - 중국 산시성(陕西省) 시안시(西安市) 명성구(明城区)를 대상으로 -

쾅바오위에,정태열,양호 한국경관학회 2024 한국경관학회지 Vol.16 No.1

도시화가 가속화되고 환경 문제가 증가함에 따라 도시 가로 녹지공간의 중요성이 점점 더 부각되고 있다. 본 연구는 딥러닝을 활용하여 스트리트 뷰(Street View Image, SVI)의 의미론적 분할을 통해 녹시율(Green View Index, GVI)을 계산하고, 원격탐사 기술을 결합하여 정규화식생지수(Normalized Difference Vegetation Index, NDVI)를 계산함으로써 서안시(西安市) 명성구(明城 区)의 가로 녹지공간의 GVI와 NDVI를 정량 분석하였다. 또한 GVI와 NDVI의 상관 관계와 공간분포를 분석하고, 다중 척도 지리 가중 회귀 모델을 사용하여 GVI와 NDVI의 공간 이질성을 더욱 탐구하였다. 연구 결과, 대부분의 가로의 GVI는 편안한 시각적 인식을 제공하는 수준에 도달하지 못했으며, 도로 등급이 낮아질수록 GVI가 감소하는 경향을 보였다. 또한 전체 녹지는 연속적인 네트워크를 형성하지 못했으며, 집중 녹지는 적고 녹지 분포가 불균등했다. GVI와 NDVI 사이에는 정적 상관 관계가 있지만, 연구 지역에는 공간적 이질성이 존재했다. 도시의 낡은 가로와 3급, 4급 도로의 녹화는 긴급히 개선이 필요하다. 연구 결과는 가로 녹지공간 계획 및 설계에참고가 될 수 있으며, 향후 도시 가로 경관 평가 및 공간 최적화를 위한 과학적 근거를 제공할수 있다. As urbanization accelerates and environmental issues multiply, the importance of urban street green spaces becomes increasingly prominent. This study utilizes deep learning for semantic segmentation of street view images (SVI) to calculate the Green View Index (GVI), and combines satellite remote sensing technology to compute the Normalized Difference Vegetation Index (NDVI). A quantitative analysis of GVI and NDVI in the green spaces along the streets of Mingcheng District in Xi’an was conducted. The correlation and spatial distribution between GVI and NDVI were analyzed, and a multiscale geographically weighted regression model was employed to explore the spatial heterogeneity between GVI and NDVI. The results indicate that the GVI of most streets does not reach the values needed for comfortable visual perception, and the GVI decreases with the lowering of road levels. Additionally, the overall green spaces have not yet formed a coherent network, with few concentrated green areas and uneven distribution of greenery. Although there is a positive correlation between GVI and NDVI, spatial heterogeneity exists in the study area. The greening of old urban blocks and level 3 and 4 urban roads urgently need improvement. The findings of this study can provide references for the planning and design of street green spaces, and offer a scientific basis for the evaluation and spatial optimization of future urban street landscapes.

A Characteristics of the Overlooking Landscape of the Original Downtown from the Traditional Landscape Architecture Space - Focuse on Viewpoints of the Overlooking Landscape of Hanyangdoseong -

최지이,이종성 한국전통조경학회 2020 한국전통조경학회지 Vol.18 No.-

This study selected and analyzed the visual characteristics of the representative view points at Hanyang-doseong(Hanyang City Wall). The results of dividing the Hanyang-doseong into four sections and examining the visual characteristics of each section's representative view points among the 32 preliminary views are as follows. First, after looking at the dips of each target landscape, the Cheongun-dae of Bukak-san and Beom-bawi of Inwang-san have a view of Gyeongbok Palace, Changdeok Palace, and Hanok Village. Naksan-jeong in Naksan is also associated with a low building group in Daehak-ro near a dip of 10°. Namsan Jamdu-bong is blocked from buildings in around 10° of dip. Second, when looking at the visual characteristics of the target landscape, the patency rate was high in the order of Cheongun-dae on Bukak-san, Beomba-wi on Inwang-san, Jamdu-bong on Nam-san, and Naksan-jeong on Nak-san. The green view index was high in the order of Cheongun-dae of Bukak-san, Beomba-wi of Inwang-san, Naksan-jeong of Nak-san, and Jamdu-bongof Nam-san. In general, the higher the view point altitude, the greater the proportion of views to the distance, and the higher the patency rate. Given that visual objects occupy a large or small view depending on the distance, even though they are of the same size, the green area also shows a difference in visual amount over distance. Third, as a result of analyzing the visual characteristics of each view point, Bukak-san and Inwang-san had a high green view index due to the high proportion of green and low-rise residential areas around the viewpoint, and the patency rate was also high. On the other hand, Nak-san and Nam-san were developed to around viewpoint, so the green view index was low. The difference between the viewpoint altitude and height of the surrounding buildings was not great, so the patency rate was also low.

전통조경공간에서 조망되는 원도심의 부감경관 특성 - 한양도성에서 바라본 부감경관을 대상으로 -

( Choi Ji-yi ),( Lee Jong-sung ) 한국전통조경학회(구 한국정원학회) 2020 Journal of Korean Institute of Traditional Landsca Vol.18 No.-

본 연구는 한양도성에서의 대표 조망점을 선정하고 이에 대한 시각특성을 분석하였다. 한양도성 구간을 4개 구간으로 구분하고, 32개 예비 조망점 중 각 구간의 대표 조망점의 시각특성을 살펴본 결과는 다음과 같다. 첫째, 대상경관별 부각 양상을 살펴본 결과 북악산 청운대와 인왕산 범바위는 부각 10° 부근의 중경이 펼쳐지며, 경복궁과 창덕궁, 한옥마을 등이 조망된다. 또한 낙산의 낙산정은 부각 10° 부근에 대학로 일대 낮은 건물군과 관계가 있다. 남산 잠두봉은 부감 영역 모두 건축물이 차단되어 있다. 둘째, 대상경관의 시각적 특성을 살펴본 결과 개방율은 북악산 청운대, 인왕산 범바위, 남산 잠두봉, 낙산의 낙산정 순서로 높게 나타났다. 녹시율은 북악산 청운대, 인왕산 범바위, 낙산의 낙산정, 남산 잠두봉 순서로 높게 나타났다. 일반적으로 조망점 표고가 높을수록 원경에 대한 조망비중이 늘면서 개방율도 비례하여 높아지며, 시각적인 대상은 같은 크기를 갖더라도 거리에 따라 크게 또는 작게 시야를 점유한다는 점을 미루어 볼 때 녹지 또한 거리에 따라 시각량의 차이를 보인다고 할 수 있다. 셋째, 각 조망점의 대상경관별 시각특성을 분석한 결과 북악산과 인왕산은 시점 주변이 녹지 및 저층주거지의 비중이 높아 녹시율이 높고, 시야가 멀리까지 확장될 조건을 갖고 있어 개방율 또한 높게 조사되었다. 반면, 낙산과 남산은 시점 주변까지 개발이 이루어져 녹시율이 낮게 조사되었으며, 시점의 표고와 주변 건축물의 높이 차가 크지 않아 개방율도 낮게 조사되었다. This study selected and analyzed the visual characteristics of the representative view points at Hanyang-doseong(Hanyang City Wall). The results of dividing the Hanyang-doseong into four sections and examining the visual characteristics of each section's representative view points among the 32 preliminary views are as follows. First, after looking at the dips of each target landscape, the Cheongun-dae of Bukak-san and Beom-bawi of Inwang-san have a view of Gyeongbok Palace, Changdeok Palace, and Hanok Village. Naksan-jeong in Naksan is also associated with a low building group in Daehak-ro near a dip of 10°. Namsan Jamdu-bong is blocked from buildings in around 10° of dip. Second, when looking at the visual characteristics of the target landscape, the patency rate was high in the order of Cheongun-dae on Bukak-san, Beomba-wi on Inwang-san, Jamdu-bong on Nam-san, and Naksan-jeong on Nak-san. The green view index was high in the order of Cheongun-dae of Bukak-san, Beomba-wi of Inwang-san, Naksan-jeong of Nak-san, and Jamdu-bongof Nam-san. In general, the higher the view point altitude, the greater the proportion of views to the distance, and the higher the patency rate. Given that visual objects occupy a large or small view depending on the distance, even though they are of the same size, the green area also shows a difference in visual amount over distance. Third, as a result of analyzing the visual characteristics of each view point, Bukak-san and Inwang-san had a high green view index due to the high proportion of green and low-rise residential areas around the viewpoint, and the patency rate was also high. On the other hand, Nak-san and Nam-san were developed to around viewpoint, so the green view index was low. The difference between the viewpoint altitude and height of the surrounding buildings was not great, so the patency rate was also low.

Analyzing green view index and green view index best path using Google street view and deep learning

Zhang Jiahao,Hu Anqi 한국CDE학회 2022 Journal of computational design and engineering Vol.9 No.5

As an important part of urban landscape research, analyzing and studying street-level greenery can increase the understanding of a city’s greenery, contributing to better urban living environment planning and design. Planning the best path of urban greenery is a means to effectively maximize the use of urban greenery, which plays a positive role in the physical and mental health of urban residents and the path planning of visitors. In this paper, we used Google street view to obtain street view images of Osaka City. The semantic segmentation model is adopted to segment the street view images and analyze the green view index (GVI) of Osaka City. Based on the GVI, we take advantage of the adjacency matrix and Floyd–Warshall algorithm to calculate GVI best path, solving the limitations of ArcGIS software. Our analysis not only allows the calculation of specific routes for the GVI best paths but also realizes the visualization and integration of neighborhood urban greenery. By summarizing all the data, we can conduct an intuitive feeling and objective analysis of the street-level greenery in the research area. Based on this, such as urban residents and visitors can maximize the available natural resources for a better life. The dataset and code are available at https://github.com/Jackieam/GVI-Best-Path.