Smart City Berlin relies on traffic data from KiwiVision

With the KiwiVision software, Berlin researchers analyze traffic in the German capital in detail without interfering with the personal rights of road users. The GRAEF Group (www.graef-gruppe.de) was responsible for the implementation. The Genetec partner specializes in consulting, planning, installation and maintenance of video, alarm, time recording and IT technology. GRAEF's customers include Siemens, Jaguar and the Charité hospital.

Unnoticed by drivers and pedestrians, work on the future of inner-city road traffic is being carried out around Berlin's Ernst-Reuther-Platz. As part of "Smart City Berlin" and subsequent projects, a consortium of companies such as T-Systems, Cisco, TÜV Nord and institutions from TU Berlin are researching the opportunities that the digitalization of the road environment offers for autonomous driving, traffic flow management and environmental protection in the metropolises of the 21st century. However, a challenge arose right from the start of the project: on the one hand, the precise recording and categorization of all road users is an indispensable basis for meaningful data models, while on the other hand, the German Data Protection Act and the European General Data Protection Regulation (GDPR) place high demands on the protection of privacy, especially in public spaces. KiwiVision has successfully reconciled these conflicting interests in technical terms.

"Most autonomous driving projects focus on vehicle-based solutions," explains Dr.-Ing. habil Manzoor Ahmed Khan, Director of Network & Mobility and one of the leading scientists at the Distributed Artificial Intelligence (DAI) Laboratory at TU Berlin. This involves equipping vehicles with sensors such as cameras, radar and lidar (distance and speed measurement by laser) as well as artificial intelligence (AI) to predict certain situations. "This approach is certainly the right one. But by digitizing the road environment and transmitting the data to the vehicle, we can significantly increase the prediction window from perhaps a hundred metres to two kilometers."

Berlin goes digital

For research purposes, numerous cameras and other sensors were therefore installed at Ernst-Reuther-Platz and along Straße des 17. Juni, providing information on road users and density, road conditions, parking space occupancy, traffic light switching and, last but not least, the weather. The immense amounts of data are then analyzed using various machine learning and other AI-based methods developed at the DAI laboratory. The potential applications of the knowledge gained are almost limitless: "In the future, we will be able to inform vehicles that the road is icy or muddy for six meters within a kilometer. However, our data models also make it possible, for example, to make precise statements about which parking spaces are likely to be free on Friday mornings and which traffic lights will favor the flow of traffic on Tuesdays at 1 pm."

The basis for such concrete statements is the correct recording of every single car, truck and bus as well as every pedestrian and cyclist. "We actually have data on all road users so that we can simulate Ernst-Reuther-Platz at all times," says Manzoor Ahmed Khan. "However, this is only possible because we paid attention to the rights of road users right from the start, whose consent we cannot obtain. We had to present a viable concept for the protection of privacy, which was rigorously checked. Otherwise we wouldn't have been given permission."

KiwiVision delivers precise, anonymized data

If the faces or license plate numbers of uninvolved road users are captured by a camera and subsequently stored, this constitutes personal data. The collection and processing of such data is limited and regulated by the GDPR, among other things. Failure to comply with the regulations can result in significant penalties. However, if people are only represented as blocks, the data does not contain any personal reference. Computer-generated representations that replace people and vehicles with symbols therefore do not fall within the scope of the GDPR.

The Berlin research team has taken advantage of this fact: KiwiVision video analysis is used with the cameras, which recognizes and categorizes objects in the traffic flow and anonymizes the data obtained. The algorithms are not fooled by different weather conditions such as snow or rain and identify moving objects just as reliably as waiting objects. Only after anonymization is the data collected by an IoT middleware developed in-house by the DAI laboratory and forwarded to processing systems. In the evaluations and simulations, trucks, cars, pedestrians etc. then appear as circles, squares or triangles - the researchers cannot and do not want to receive any more information. Subsequent de-anonymization is impossible, as there are no recordings of the video streams. "Because recording and identification is so central to our project, we naturally looked at several technologies and carried out comparative studies," says Manzoor Ahmed Khan. "KiwiVision definitely met all our requirements the best, so we decided to use it, which proved to be the right decision later on."

The same algorithms are also used in the KiwiVision Privacy Protector from Genetec. With this pixelation and masking solution, companies can ensure that employees and visitors are rendered unrecognizable during video surveillance. The Privacy Protector fully integrates with Genetec Security Center, the centralized platform for unified management of the entire physical security infrastructure. By supporting open industry standards, Genetec Security Center enables the management of all non-proprietary IP-based cameras, door locks, sensors, etc. via a single interface.

In contrast to the Berlin research project, the original video stream is recorded in encrypted form when used by the company and is available for investigating incidents. Access is secured by an authentication procedure that requires two employees to enter special chip cards and PINs or login information. The Privacy Protector is the only solution to have been awarded the European Privacy Seal. This seal of approval is only awarded after intensive technical and legal testing of the compatibility of IT products with European data protection guidelines. The seal is valid for two years so that certified products always meet the latest requirements.

Further use of KiwiVision

Smarter traffic management is one of the great challenges of our time, which is why the work of the Berlin researchers is far from over. Another large-scale project on 5G mobile communications and autonomous driving is already planned, in which Ernst-Reuther-Platz and Straße des 17. Juni will form one of six European test routes. KiwiVision video analysis will once again provide the necessary information for this project. Manzoor Ahmed Khan comments: "KiwiVision is our solution for anonymized traffic data. The support was very cooperative at all times. Based on our previous experience, we are therefore looking forward to working together on further projects."