2:30 PM - 2:45 PM
[HTT14-09] The generation process and data characteristics of volunteered street view imagery for streetscape monitoring: a case study in Tokyo
★Invited Papers
Keywords:crowdsourcing, Volunteered Street View Imagery, Mapillary, contribution behavior
The surge of Street View Imagery (SVI) as an essential data source for urban analytics, especially in streetscape audit studies, has been catalyzed by the proliferation of imagery platforms, advances in computer vision, machine learning, and the availability of computing resources. However, it has been noted that the mainstream data provided by government agencies and private companies (e.g., Google Street View, Baidu Total View) have limitations in spatial resolution, update frequency (e.g., GSV updates data every few years), and data application (e.g., barriers to free data download, historical data retrieval, and free data use). These restrictions present issues when achieving more accurate and dynamic streetscape monitoring.
The emergence of the Web 2.0 era has fostered the potential for individuals to contribute and access information through multiple resources, which has also facilitated the collection of massive Volunteered Street View Imagery (VSVI). The VSVI data have the potential to provide more open, comprehensive, and diverse geographic information, which is however conditional on a set of criteria such as data completeness and quality. To better understand the value of this novel type of data in streetscape monitoring studies, this study aims to analyze the generation process of VSVI and examine relevant characteristics in Tokyo using the typical VSVI data of Mapillary.
The generation process of Mapillary from 2014 (the inception of Mapillary) to 2022 is analyzed from the perspective of road expansion, data amount accumulation, and hotspot change of VSVI data; the examination of characteristics includes the assessment of spatial distribution (road coverage and spatial density), contribution time distribution (revisit time, update frequency, and seasonal diversity), and image quality (image type and shooting perspective). These analyses use GSV imagery data as a benchmark.
The emergence of the Web 2.0 era has fostered the potential for individuals to contribute and access information through multiple resources, which has also facilitated the collection of massive Volunteered Street View Imagery (VSVI). The VSVI data have the potential to provide more open, comprehensive, and diverse geographic information, which is however conditional on a set of criteria such as data completeness and quality. To better understand the value of this novel type of data in streetscape monitoring studies, this study aims to analyze the generation process of VSVI and examine relevant characteristics in Tokyo using the typical VSVI data of Mapillary.
The generation process of Mapillary from 2014 (the inception of Mapillary) to 2022 is analyzed from the perspective of road expansion, data amount accumulation, and hotspot change of VSVI data; the examination of characteristics includes the assessment of spatial distribution (road coverage and spatial density), contribution time distribution (revisit time, update frequency, and seasonal diversity), and image quality (image type and shooting perspective). These analyses use GSV imagery data as a benchmark.