A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Green streets − Quantifying and mapping urban trees with street-level imagery and computer vision
https://orcid.org/0000-0002-9188-9225
HighlightsCurrent methods to map urban trees are hindered by high costs or poor resolution.Computer vision is used to quantify urban tree cover from street view images.A multi-step algorithm accurately segments the percent of tree cover in an image.Modeling the relationship between neighbouring images increases predictive accuracy.Computer vision can map urban tree cover at high spatial resolution with rapid throughput.
AbstractTraditional tools to map the distribution of urban green space have been hindered by either high cost and labour inputs or poor spatial resolution given the complex spatial structure of urban landscapes. What’s more, those tools do not observe the urban landscape from a perspective in which citizens experience a city. We test a novel application of computer vision to quantify urban tree cover at the street-level. We do so by utilizing the open-source image data of city streetscapes that is now abundant (Google Street View). We show that a multi-step computer vision algorithm segments and quantifies the percent of tree cover in streetscape images to a high degree of precision. By then modelling the relationship between neighbouring images along city street segments, we are able to extend this image representation and estimate the amount of perceived tree cover in city streetscapes to a relatively high level of accuracy for an entire city. Though not a replacement for high resolution remote sensing (e.g., aerial LiDAR) or intensive field surveys, the method provides a new multi-feature metric of urban tree cover that quantifies tree presence and distribution from the same viewpoint in which citizens experience and see the urban landscape.
Green streets − Quantifying and mapping urban trees with street-level imagery and computer vision
https://orcid.org/0000-0002-9188-9225
HighlightsCurrent methods to map urban trees are hindered by high costs or poor resolution.Computer vision is used to quantify urban tree cover from street view images.A multi-step algorithm accurately segments the percent of tree cover in an image.Modeling the relationship between neighbouring images increases predictive accuracy.Computer vision can map urban tree cover at high spatial resolution with rapid throughput.
AbstractTraditional tools to map the distribution of urban green space have been hindered by either high cost and labour inputs or poor spatial resolution given the complex spatial structure of urban landscapes. What’s more, those tools do not observe the urban landscape from a perspective in which citizens experience a city. We test a novel application of computer vision to quantify urban tree cover at the street-level. We do so by utilizing the open-source image data of city streetscapes that is now abundant (Google Street View). We show that a multi-step computer vision algorithm segments and quantifies the percent of tree cover in streetscape images to a high degree of precision. By then modelling the relationship between neighbouring images along city street segments, we are able to extend this image representation and estimate the amount of perceived tree cover in city streetscapes to a relatively high level of accuracy for an entire city. Though not a replacement for high resolution remote sensing (e.g., aerial LiDAR) or intensive field surveys, the method provides a new multi-feature metric of urban tree cover that quantifies tree presence and distribution from the same viewpoint in which citizens experience and see the urban landscape.
Green streets − Quantifying and mapping urban trees with street-level imagery and computer vision
https://orcid.org/0000-0002-9188-9225
HighlightsCurrent methods to map urban trees are hindered by high costs or poor resolution.Computer vision is used to quantify urban tree cover from street view images.A multi-step algorithm accurately segments the percent of tree cover in an image.Modeling the relationship between neighbouring images increases predictive accuracy.Computer vision can map urban tree cover at high spatial resolution with rapid throughput.
AbstractTraditional tools to map the distribution of urban green space have been hindered by either high cost and labour inputs or poor spatial resolution given the complex spatial structure of urban landscapes. What’s more, those tools do not observe the urban landscape from a perspective in which citizens experience a city. We test a novel application of computer vision to quantify urban tree cover at the street-level. We do so by utilizing the open-source image data of city streetscapes that is now abundant (Google Street View). We show that a multi-step computer vision algorithm segments and quantifies the percent of tree cover in streetscape images to a high degree of precision. By then modelling the relationship between neighbouring images along city street segments, we are able to extend this image representation and estimate the amount of perceived tree cover in city streetscapes to a relatively high level of accuracy for an entire city. Though not a replacement for high resolution remote sensing (e.g., aerial LiDAR) or intensive field surveys, the method provides a new multi-feature metric of urban tree cover that quantifies tree presence and distribution from the same viewpoint in which citizens experience and see the urban landscape.
Green streets − Quantifying and mapping urban trees with street-level imagery and computer vision
Seiferling, Ian (author) / Naik, Nikhil (author) / Ratti, Carlo (author) / Proulx, Raphäel (author)
Landscape and Urban Planning ; 165 ; 93-101
2017-05-13
9 pages
Article (Journal)
Electronic Resource
English
From “streets for traffic” to “streets for people”: can street experiments transform urban mobility?
| Taylor & Francis Verlag | 2020
Successful streets : performance measures, community engagement, and urban street design
| DSpace@MIT | 2012