딥러닝을 활용한 거리뷰 기반 도시 범죄요인 분석:범죄예방 환경설계(CPTED) 전략을 중심으로

Crime is a significant social issue in modern cities, undermining citizens' safety and quality of life, making research on physical environmental factors essential for addressing this problem. Existing studies have faced limitations in objectively and quantitatively analyzing these factors. Therefore, this study aims to analyze the impact of urban physical environments on crime occurrence by utilizing street view images and deep learning techniques. Specifically, based on CPTED (Crime Prevention Through Environmental Design) principles, which focus on crime prevention through physical environment design, relevant environmental factors were selected and quantitatively evaluated. The deep learning model DeepLabV3+ was applied to extract physical environmental factors from street view images in the Boston area, and the spatial distribution of these factors was analyzed. Additionally, Spearman correlation analysis with actual crime data was conducted to determine the impact of each factor on crime occurrence. The results showed significant correlations between crime occurrence and factors related to 'Territoriality' and 'Activity Support' in CPTED principles, with greenery in particular playing an important role in crime occurrence. Through these findings, this study is expected to contribute to the creation of safe and sustainable urban environments and inform crime prevention policies by identifying key factors that influence crime.