Intelligent video surveillance with more privacy
Intelligent video analysis methods are capable of combining increased security with improved protection of personal rights. This is proven by the latest developments.
Germany's new police law allows intelligent video surveillance in the future for preventive police purposes in three cases: at crime hotspots, at "endangered objects" and at public events if there is a threat of terrorist attacks. The intelligent technology is to be tested and further developed in a pilot project in Mannheim in parallel with the existing video surveillance.
"Technically enforcing the required data protection".
"Privacy is an important concern for us in video evaluation, among other things," emphasized Jürgen Beyerer, director of the Fraunhofer Institute of Optronics, System Technologies and Image Exploitation (IOSB). "Our approach is to technically enforce the required data protection and protection of personal rights. Only in this way, we are convinced, will such systems find acceptance."
Markus Müller, spokesman for the "Civil Security" business unit at Fraunhofer IOSB, explained how this works: While conventional surveillance technology records and stores every person in the highest resolution, even if they are going about harmless everyday business, intelligent video surveillance offers the possibility of "cascaded anonymization": "In this process, irrelevant areas, scenes and people are pixelated by the technology, for example, or faded out completely." Only when an algorithm detects something suspicious does the system focus the image and alert the human operator. Says Müller, "We have processes in development for this, for example, to detect assaults or parked and orphaned objects." The software is action-based. In other words, it is designed to recognize action patterns.
Automatic search for specific characteristics
In addition to this preventive aspect, intelligent video evaluation processes can also provide important support for subsequent investigative work, provided that the legal basis for authorization exists in the Code of Criminal Procedure. As an example, Müller cited the recognition of so-called soft biometric features such as certain accessories, hair color or body size. "Such characteristics are for the most part changeable and not suitable for identification due to their low discriminatory ability." Nevertheless, automated searches for soft biometric features in video data could help investigators after a crime has been committed, he said. "In many cases today, police officers often spend endless hours sitting in front of a screen reviewing countless videos. Our systems can massively facilitate and shorten this tedious work for police officers."
Fraunhofer IOSB has been working for years on the development and continuous improvement of pattern-based algorithms for automatic searches in image and video data. Corresponding software is already in use at some state criminal investigation departments.
* Ulrich Pontes, Head of Public Relations Fraunhofer Institute of Optronics, System Technologies and Image Exploitation IOSB, Karlsruhe, Germany
