Graffiti Detection


Traditional CCTV surveillance systems for detecting vandalism require constant human operator-based monitoring and intervention. In this project, the implemented prototype system (mounted at a bus station) can successfully detect graffiti and graffiti-making activities through remove noise signals introduced by passing pedestrians and vehicles as well as adapting to changing outdoor illumination conditions.

One common assumption for graffiti detection is based on its time-endurance feature due to the adhesive paint coating. Therefore, the most convenient and effective real-time solution for graffiti detection is to compare a pre-determined “clean” scene with the live feeds from an image or video device.

system freamwork

The Implementation of Human Noise Removal Based-on Head Curve Matching
Compared with the complex task of modelling an entire human body, defining and tracking a human head is a much simpler job since it is a relatively rigid part of the human body. More importantly, all people seem to show an identical head and shoulder contour, an “Ω” shape, even when being oriented to different directions

  • The main contribution of this project is the development of a rapid technique for removing foreground noise caused by passengers  painting at a bus stop through head curve matching.
  • Test results show the algorithm and process developed in the system prototype is flexible and robust, which can be adopted for handling other human contour-based detection tasks.
  • As part of the future work for this research, human gestures will be studied based on human stick-models for real-time gesture analysis to identify other forms of vandalism acts.
Graffiti detection: Automated crime detection in security camera feeds

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>