kth.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Crime and Visually Perceived Safety of the Built Environment: A Deep Learning Approach
KTH, School of Architecture and the Built Environment (ABE), Urban Planning and Environment. Department of Urban Planning and Environment and Senseable Stockholm Lab.ORCID iD: 0000-0003-2050-8365
GISense Lab, Department of Geography and the Environment, University of Texas at Austin.ORCID iD: 0000-0003-3810-9450
KTH, School of Architecture and the Built Environment (ABE), Urban Planning and Environment, Urban and Regional Studies.ORCID iD: 0000-0001-5302-1698
Department of Urban Planning and Environment and Senseable Stockholm Lab, KTH Royal Institute of Technology.ORCID iD: 0000-0001-7606-8771
Show others and affiliations
2025 (English)In: Annals of the American Association of Geographers, ISSN 2469-4452, E-ISSN 2469-4460, p. 1-21Article in journal (Refereed) Published
Abstract [en]

Although the influence of the built environment on both crime and people’s safety perceptions is well documented in the international literature, less evidence is found regarding the link between urban safety perceptions and crime occurrence. In this article, we investigate the potential relationship between crime and visual perceived safety (VPS), using Stockholm, Sweden as a case. Central to the study is the VPS score, a detailed measure of VPS and situational fear, created by combining a deep learning model with a data set of local street view images and citizen impressions. We examine this measure together with traditional crime records to compare the city’s distribution of safety and crime. First, geographical patterns and spatial clusters of high and low levels of crime and VPS were detected. Then, drawing from principles of environmental criminology, a spatial regression was used to examine the relationship between the VPS score and crime, controlling for sociodemographics and land-use factors. Findings show that crime rates of different types are significant predictors of poor VPS, but mismatching geographies of perceived safety and crime are common. The article discusses the findings and finishes by highlighting the impact of these results for research and practice.

Place, publisher, year, edition, pages
Informa UK Limited , 2025. p. 1-21
National Category
Criminology
Identifiers
URN: urn:nbn:se:kth:diva-364194DOI: 10.1080/24694452.2025.2501998ISI: 001497335300001Scopus ID: 2-s2.0-105006974047OAI: oai:DiVA.org:kth-364194DiVA, id: diva2:1964675
Note

QC 20250609

Available from: 2025-06-05 Created: 2025-06-05 Last updated: 2025-07-04Bibliographically approved
In thesis
1. To Catch Fear In A Bottle: Conceptualizing and Measuring Perceived Safety in Urban Environments
Open this publication in new window or tab >>To Catch Fear In A Bottle: Conceptualizing and Measuring Perceived Safety in Urban Environments
2025 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

Cities are complex, dynamic and constantly evolving systems, shaped by a plethora of factors that in turn significantly impact our individual safety. These factors are embedded in everything from the physical design of urban settings to the social interactions and activities that take place within them. Compared to other urban challenges such as crime, perceived safety and fear are more abstract and subjective phenomena that are much more difficult to define, understand and not to mention prevent. The aim of this thesis is to examine how perceived safety can be conceptualized and measured in urban environments in order to enhance the accuracy and comprehensiveness of urban safety diagnostics. The approaches of the studies explore both more traditional methods of surveys, regression models and spatial analysis, as well as more emergent methodologiesof the past few decades, such as deep learning models and mobile app data collection. The studies include analyses of perceived safety on multiple levels, from meso- to microlevel, comparisons of residents and non-residents, as well as intersectional analyses of gender, age, ethnicity, and income. The findings of the thesis highlight how various conceptualizations of perceived safety and methodologies used to measure them reveal different understandings. Local and global measures of fear have inverted relationships with marginalized men in Sweden, who, in contrast to common beliefs, express similar levels of poor neighborhood safety to women overall. Spatial and conceptual contrasts were identified between survey-based measures of neighborhood safety, objective safety measures such as crime rates, and safety scores generated by deep learning models. A survey based on a public participation geographic information system (PPGIS) revealed the moderating effect of place familiarity and place stigma, where minor differences in neighborhood familiarity were found to significantly affect perceptions. Mobile app data collection was also found to be a potential way to capture highly spatio-temporally detailed data on perceived safety, while retaining the ability to capture personal narratives and emotional connections to place. The thesis concludes with a reflection of implications and practical insights of the findings for researchers, urban planners, and policymakers, encouraging strategies for more comprehensive diagnostics of safety. 

Abstract [sv]

Städer är komplexa, dynamiska och ständigt föränderliga system, formade av en mängd olika faktorer som i sin tur har en betydande inverkan på vår individuella trygghet. Dessa faktorer är inbäddade i allt från den fysiska utformningen av urbana miljöer till de sociala interaktioner och aktiviteter som äger rum inom dem. Jämfört med andra urbana utmaningar såsom brottslighet är upplevd trygghet och rädsla mer abstrakta och subjektiva fenomen som är mycket svårare att definiera, förstå och inte minst att förebygga. Syftet med denna avhandling är att undersöka hur upplevd trygghet kan konceptualiseras samt mätas i urbana miljöer för att förbättra träffsäkerheten och omfattningen av trygghetsdiagnostik i städer. Angreppssätten i de inkluderade studierna omfattar både mer traditionella metoder såsom enkäter, regressionsmodeller och geografiska analyser, men även nyare metoder från de senaste decennierna såsom djupinlärningsmodeller (deep learning-modeller) och datainsamling via mobilapplikationer. Studierna inkluderar analyser av trygghet på flera nivåer, från meso- till mikronivå, jämförelser mellan boende och icke-boende, samt intersektionella analyser av kön, ålder, etnicitet och inkomst. Avhandlingens resultat belyser hur olika konceptualiseringar av upplevd trygghet och de metoder som används för att mäta dem ger upphov till olika förståelser. Lokala och globala mått på rädsla visade omvända samband för marginaliserade män i Sverige, vilka uttrycker – i motsats till rådande uppfattningar – liknande nivåer av otrygghet i sina bostadsområden som kvinnor i stort. Geografiska och konceptuella kontraster identifierades mellan enkätbaserade mått på trygghet i bostadsområden, objektiva säkerhetsmått såsom brottsstatistik, samt trygghetspoäng genererade av djupinlärningsmodeller. En enkät baserad på PPGIS (Participatory Public Geographic Information System) visade på den modererande effekten av platsbekantskap och platsrelaterad stigma, där små skillnader i kännedom om området hade stor påverkan på trygghetsuppfattningen. Datainsamling med hjälp av en mobilapplikation visade också potential att fånga in mycket detaljrik rumslig och tidsmässig data om upplevd trygghet, samtidigt som den möjliggör insamling av personliga berättelser och känslomässiga band till platsen. Avhandlingen avslutas med en reflektion kring implikationer och praktiska insikter av resultaten för forskare, stadsplanerare och beslutsfattare, med en uppmaning till strategier för mer heltäckande trygghetsdiagnostik.

Place, publisher, year, edition, pages
Stockholm: KTH Royal Institute of Technology, 2025. p. 66
Series
TRITA-ABE-DLT ; 2517
Keywords
Urban safety, spatial analysis, GeoAI, territorial stigma, participatory mapping, Urban trygghet, rumslig analys, GeoAI, territorial stigma, deltagarstyrd kartläggning
National Category
Criminology Human Geography
Research subject
Urban and Regional Planning
Identifiers
urn:nbn:se:kth:diva-364197 (URN)978-91-8106-326-4 (ISBN)
Public defence
2025-08-28, Kollegiesalen, Brinellvägen 8, KTH Campus, public videoconference link https://kth-se.zoom.us/j/61193674132, Stockholm, 09:00 (English)
Opponent
Supervisors
Note

Research funders: Stockholm Senseable Lab, Stockholm Chamber of Cimmerce, Newsec, Digital Futures

QC 20250627

Available from: 2025-06-27 Created: 2025-06-06 Last updated: 2025-07-01Bibliographically approved

Open Access in DiVA

No full text in DiVA

Other links

Publisher's full textScopus

Authority records

Abraham, JonatanCeccato, Vania

Search in DiVA

By author/editor
Abraham, JonatanKang, YuhaoCeccato, VaniaNäsman, PerDuarte, FábioGao, SongLjungqvist, LukasZhang, Fan
By organisation
Urban Planning and EnvironmentUrban and Regional Studies
In the same journal
Annals of the American Association of Geographers
Criminology

Search outside of DiVA

GoogleGoogle Scholar

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 24 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf