| Title |
Developing Geo-coded Street-level Pedestrian Volume Data Using Google Street View Data and Artificial Intelligence Models |
| Authors |
김영우(Kim, Youngwoo) ; 황용하(Hwang, Yongha) ; 정은석(Jeong, Eunseok) ; 강범준(Kang, Bumjoon) |
| DOI |
https://doi.org/10.5659/JAIK.2023.39.9.57 |
| Keywords |
Street Activities; Street Data; Pedestrian Volume; Image Detection; Artificial Intelligence |
| Abstract |
Pedestrian count data serves various purposes within architectural, urban planning, and related fields. Typically, this data is collected by
government agencies and commercial survey companies. However, conventional methods of recording pedestrian data demand significant time
and effort. Consequently, data availability is restricted to specific timeframes and limited locations. In response to this, we conducted
feasibility tests for an object-based pedestrian detection procedure. Google Street View data was used to capture geocoded pedestrian counts at
street levels in New York City, the U.S. A validation study was performed against historical pedestrian count data recorded officially in the
city at 114 different locations. The results indicated a high agreement rate of over 0.8, suggesting that street-level image data could
effectively and economically replace conventional pedestrian counting methods. |