top of page

Eyes on the Street

  • Nov 12
  • 2 min read
ree

Natural-surveillance analyses street-view images to measure how visible streets are from nearby buildings, indicating safety and observation levels.


Purpose of Tool:

Natural-surveillance is an open-source Python methodology for estimating “natural surveillance” (i.e., the passive “eyes on the street” effect) at scale by analysing street-level imagery with computer vision techniques. It detects building openings (e.g., windows, doors) in street-view imagery, localizes them in 3 D relative to street segments, and computes metrics for street segments such as “road surveillability” and “occupant surveillability”. The method was applied to neighbourhoods in Amsterdam to examine the relationship between these metrics and actual/perceived safety. GitHub+1In the Equal-Life context, this tool helps quantify built-environment exposure related to safety and surveillance which may influence children’s mobility, independent outdoor activity, social interactions and mental health (via the physical and social exposome). It adds a measurement layer to the built-environment exposure chain by providing metrics for how the built form supports or inhibits passive observation and hence potentially affects safety, social cohesion, and environmental stressors.


Classification of tool:

Model / analytical geospatial tool (Python library for built-environment visibility & surveillance metrics).


Required skills:

  • Python programming (to run the pipeline, configure settings, process imagery) GitHub

  • GIS / geospatial data handling (street-segment mapping, coordinate systems, sight-line geometry) repository.tudelft.nl

  • Computer-vision fundamentals (image processing, facade labelling, deep learning for window detection) GitHub


Required input data:

  • Street-level imagery (e.g., from Google Street View Static API) covering the area of interest. GitHub+1

  • Building facade data / extracted window/door openings detected via computer vision. GitHub

  • Street-segment network with geometry and reference coordinates (so that surveillance metrics can be tied to street segments).

  • Parameters for sight‐line distance, viewing angle, field-of-view (configurable via the pipeline) GitHub


Output:

  • A dataset of street segments (or blocks) each annotated with natural surveillance metrics (e.g., road surveillability score, occupant surveillability score) repository.tudelft.nl

  • Geospatial data files (such as shapefiles or geodataframes) for mapping or further analysis of surveillance across urban space.

  • Statistical relation outputs (e.g., correlation of surveillance metrics with safety indices or crime data) as produced in the demonstration in Amsterdam. repository.tudelft.nl


Relation to other tools:

  • It complements other built-environment analysis tools (such as network accessibility tools like CTnetwork or walkability/co-accessibility tools like coaccessibility) by focusing on the visibility/surveillance dimension of urban form—an important but sometimes overlooked facet of the built‐environment exposure.

  • It can feed into exposure modelling frameworks within Equal-Life by adding a metric for how “observed” or “passively surveilled” a child’s neighbourhood might be, which could influence outdoor mobility, play, physical activity and social interaction exposures.

  • It supports the overall aim of the Toolbox for open-source, scalable methods to quantify exposures in the external/physical exposome (WP3) and potentially intersecting social-exposome elements (WP4) by operationalising safety/visibility as metrics.

  • The output of natural-surveillance can be combined with other exposure layers (e.g., access to green space, pedestrian connectivity, noise, social deprivation) to build richer multi-factor models for children’s development and mental health outcomes.


Links:

Comments

bottom of page