Image and Graffiti Recognition

Analyses photos of graffiti incidents using machine learning to map and track graffiti offending across Christchurch.

The graffiti recognition app tracks graffiti activity using data from the public and contractors to help curb damage and identify repeat offenders. We analyse each photo using the newly developed graffiti recognition software, and capture key data from the images to help identity and map the activities of recidivist offenders. The data collected and visualised by the solution is used by Council staff, community groups, and NZ Police, to inform decision-making that will reduce graffiti incidents.

Smart Christchurch has worked with Flock Consulting as the delivery partner for the graffiti recognition project. The platform matches graffiti images with other images most similar that have been collected and stored in the application. This enables identification of the most prolific taggers and related tags, creating groups with associated ID’s to be used for later reference.  This information can then be utilised by relevant stakeholders.

We explored different feature extraction algorithms and machine learning approaches, then used the images/training data to test models and develop a user-friendly front-end web app to:

  • present a series of images ranked from most to least similar
  • suggest matches where sufficient tagged data exists already
  • allow the user to quickly select images of the ‘obvious’ match
  • create a group for the purposes of tagger identification

Background

The Council receives approximately 20,000 notifications of graffiti incidents from the public each year. The majority come via the ‘Snap Send Solve’ platform, as well as emails and online forms. Investigation done in partnership with the Council’s Graffiti Programme and the Continuous Improvement team highlighted an opportunity to work with an innovative technology partner to develop the capability to analyse images received each year, and capture key data for each incident using machine learning.

Developing an automated graffiti-analysis solution using machine-based learning in a joint partnership with an external vendor is a cost effective and innovative way to counter what is a costly, growing problem for Christchurch.

Council can now use the information captured to map and track graffiti offenders operating across Christchurch. This information is also shared with the police to help build case files on offenders and enable prosecutions to be pursued. As this this is further developed, there will be an opportunity in the future to share and/or license the solution to other authorities in New Zealand and overseas.

Expected outcomes

  • Develop the capability to analyse and correlate photos of graffiti incidents using machine learning
  • Ensure the data gathered though this trail is considered reliable by all stakeholders
  • Map and track graffiti offender activity across Christchurch
  • Visualise the data submitted by the public in an accessible and easily understood real-time process
  • Build case files for recidivist offenders that other agencies such as NZ Police and Corrections can leverage
  • Provide a real and effective deterrent for one-off and recidivist offenders
  • Ensure the solutions used in these trials are endorsed by CCC and external partners and implemented as business as usual (BAU)
  • Reduce the cost burden to ratepayers of removing graffiti through drop in offending rates
  • Understand the activity levels and locations of individuals and ‘crews’ for pre-emptive and/or covert monitoring
  • The developed solution will be shared or licensed to other authorities in New Zealand and overseas

Progress

Shortlisted for 2021 LGNZ Excellence awards in the Social Wellbeing category

Community outcomes

Resilient communities | Liveable city