Street Visual Environment And Criminal Behavior Modeling

Streets are the space carrier for residents’ daily social activities.On the one hand,it undertakes diversified services functions such as residence,commerce,transportation,and leisure.On the other hand,highly open streets are more likely to become crime prone areas.Street visual environment elements can affect people’s psychology and behavior,which has an important impact on the occurrence and evolution of criminal activities.The in-depth study of the influence mechanism of street visual environment on crime and discovering the key street environmental factors that affect criminal behavior can provide important reference and guidance for future crime prevention and control,street management and street design.However,due to the limitation of street environment data acquisition,the quantitative research on street environment and criminal behavior is relatively lacking,resulting in the influence mechanism between street environment and criminal behavior is still unclear.Therefore,how to automatically and quickly measure visual environment on the street scale based on the human eye perspective,quantitatively model the relationship between criminal activities and the street visual environment,and reveal the hidden driving mechanism is a very meaningful scientific problem.Recently,the rise of street view images and the development of computer vision technology have brought new ideas for the quantification of street physical environment,but also brought new opportunities for quantitative and qualitative research on the relationship between street environment and criminal behavior.To this end,this thesis uses Google Street View(GSV)images to explore the relationship between street visual environment elements and criminal behavior.This thesis firstly carries out a quantitative measurement on the collected GSV images.Specifically,deep learning technology is used to perform semantic segmentation on GSV images,propose and calculate eight types of visual indicators that may have an impact on crime,and analyze the spatial and quantitative distribution of each indicator.Then,it also analyzes the temporal and spatial distribution of criminal behavior,uses the hot spot analysis method to identify cold and hot spots of different crime types,and combines visual indicators to preliminarily explore the relationship between crime and the street environment.Next,using these eight indicators as explanatory variables,and the number of different crime types as dependent variables,the poisson regression model and the geographically weighted poisson regression(GWPR)model are constructed respectively to analyze the relationship between street view indicators and different crime types from different perspectives.An experiment was conducted in downtown and uptown Manhattan,New York.The results show that:1)There are obvious differences in the spatial and quantitative distribution of street visual indicators between uptown and downtown.Particularly,the Green View Index(GVI)value of uptown is larger,and the uptown is visually more "green" than downtown.2)The distribution of crimes on the road segments is spatially clustered,and a large number of crimes are gathered on a small number of road segments.In addition,violent crimes are clearly concentrated in the streets of Harlem in uptown and the lower east city in downtown.3)From a global perspective,the Motorization Index(MI)has a significant role in promoting property crime,violent crime,daytime crime and nighttime crime,while the impact of GVI and Light View Index(LVI)on crime is largely dependent on socio-economic factors.4)From a local perspective,the relationship and intensity between the Pedestrian Space Index(PSI),GVI,the LVI,the Building View Index(BVI),MI and criminal activities have changed significantly in space,and LVI has the most dramatic changes in space.BVI has a strong inhibitory effect on property crime and violent crime,and LVI has a strong inhibitory effect on property crime and night crime.5)Affected by geographical location,social economy,customs and culture,and differences in the combination of street physical elements,even on streets with the same visual indicators,the impact of visual indicators on crime may be different.6)Based on the strength of the above-mentioned influence relationship,the key street visual physical elements that affect specific criminal activities on any street can be found.