생활도로 노상주차 식별을 위한 Google Street View API와 딥러닝 모형의 적용

초록

This study has trained and validated a deep learning model that identifies street parking on the streets of Seoul by utilizing the street view images obtained through the Google Street View API. We derived the final model which shows highest accuracy of object detection among models with variation of the number of training iterations. The final model shows 75.68% accuracy of object detection for all object types and 82.07% for street parking vehicles. The main implications of this study are as follows. This study introduces improved data collection method of street parking data in terms of time and monetary cost compared to the conventional field survey. Also this study applies the deep learning using street view image on the collection of spatial information for the urban planning and spatial analysis field.

키워드