Experts from FIT developed a surveillance camera with automatic calibration. The co-author received a prestigious award

Vehicle speed measurement, automatic vehicle recognition and estimation of their time of arrival, or analysis of free parking spaces number through surveillance cameras. These are just some of the projects in which Jakub Sochor from the Faculty of Information Technology of BUT participated. Last year, he was awarded the Joseph Fourier Prize for his research in computer science and informatics. He took over this prestigious award directly from a Nobel Prize winner at the French embassy.

Thanks to the research conducted by Jakub Sochor from FIT BUT, a single surveillance camera located at crossroads and on roads could be sufficient for monitoring of all lanes in the future. FIT experts have developed a system that allows any camera to calibrate automatically. "It is possible to use an ordinary camera that can be mounted on a bridge, for example. It calibrates itself and finds out everything about the road. In addition, it can be zoomed out. We can scan three lanes at once using a single camera," said Sochor adding that at the moment there must be a surveillance camera for each lane and each direction to scan multilane roads.

During the research, researchers also tested the possibility of measuring speed using a surveillance camera. "We got data with an average error of 1.10 kilometres per hour, which is a very good result. We also compared our measurements with radars and the results are very similar," added Sochor. But drivers do not have to worry. The measured values were not sent anywhere. In the future, however, police could use such devices. At the same time, the data can also be used for statistical purposes to determine the average speed of drivers in a given stretch. "We are currently discussing co-operation and possible application of the research with a company based in Brno. However, it is still in its early stage," says Jakub Sochor.

The Faculty of Information Technology of BUT is continuing with the research in traffic surveillance. Another project Jakub Sochor worked on is vehicle re-identification. "It's an automatic vehicle recognition. Vehicle recognition is currently done by identifying and recording license plates. Our system is anonymous and is based on calculating of feature vectors for individual cars," explained Jakub Sochor. Specifically, the system can detect a passing car and calculate a feature vector, which is a set of numbers. As soon as the car passes under the following surveillance camera, the system recalculates the feature vector and compares it to those previously recorded. "It is then able to evaluate with relatively high accuracy whether or not it is still the same car," added Jakub Sochor.

Currently, the success rate on the short stretches is over sixty percent. But the problem arises when two identical car models of the same colour are driving near one another. "In such a case, the system may be mistaken," he confirmed. Car detection can be used to estimate times of arrival. For example, truck operators would know how long it takes to drive along the chosen routes. Since January, Jakub Sochor has also been involved in a project of the Technology Agency of the Czech Republic, which deals with the use of cameras in parking space monitoring. "We are devising a system that would automatically analyse and evaluate the number of free parking spaces to make parking faster and more efficient," Sochor added.

The young researcher was awarded the Joseph Fourier Prize for his research activities in computer science and informatics last year. He was nominated by his faculty and won the third place for the presentation of his work. He was awarded the prize at the French Embassy by the Nobel Prize in Chemistry laureate Jean-Marie Lehn.