The internet offers an attractive platform to connect geographic locations with information of persons or companies of the real world. Some recently developed web based applications present interactive viewable pictures of urban environment to the user, where buildings and sights are complemented with informative content to e.g. inform about the history of sights.
Such technology offer completely new business models for the provider and the customer or client. In addition to Google Street View you can find some examples at www.sightwalk.de of the Cologne-based projectpartner panogate GmbH or other providers.
The enacted data protection regulations demand to anonymize each and every shown image content, though. Especially there must not be any recognizable faces or license plates published in the internet. A considerable amount of the process for the integration of new pictures falls to the anonymization of such images, which has usually to be done manually for every image. With a data volume of some tenthousand images for the core area of a large city, a manual annonymizing process is going to be time consuming and costly, represents a monotonous workload and eventually slows the development of the product down.
The goal of this project is to develop a computer based algorithm for anonymization, based on latest methods of computer vision, for the presented causes, which reliably recognizes faces and license plates, even in uneven illuminated pictures and in variant distances or physical orientations, and afterwards anonymize those in accordance with data protection regulations.