This action is co-financed by the European Union from the European Regional Development Fund The contents of this poster are the sole responsibility of the University of Zagreb, Faculty of Electrical Engineering and Computing and do not necessary reflect the views of the European Union. VISTA – Computer Vision Innovations for Safe Traffic Research activities Expected impacts imp The impacts on the final beneficiaries are: Development of automotive industry sector and higher employment. Safer traffic for drivers and pedestrians. Reduced environmental pollution. Reduced expenses for maintenance and fixing of traffic accident-related damages. Funding Science and Innovation Investment Fund Grant Scheme DurationApril 2013 – March 2015 Budget€ ,45 EU contribution€ ,29 Coordinator University of Zagreb, Faculty of Electrical Engineering and Computing Partner University of Zagreb, Faculty of Transport and Traffic Sciences Automatic white balance and brightness adjustment Lane detection and recognition The developed methods for lane detection use stereoscopic 3D structure and motion estimation in order to improve the performance. 3D reconstruction can be also used for obstacle detection. Lane detection and obstacle detection from stereo reconstruction Traffic sign detection and recognition Traffic sign recognition system can be used for driver assistance by reminding the driver or generating a warning if necessary. It can be also used for automated inspection of the state of traffic infrastructure, or as an additional source of information for autonomous navigation. Traffic sign detection and recognition Moving objects detection for collision warning Development of robust algorithms for detecting moving objects based on laser range sensor scans or omnidirectional camera acquisition. The results can be used in to inform the driver about potential collisions. Automatic driver fatigue monitoring The aim of this task is to develop a reliable real-time system for monitoring driver fatigue and drowsiness in order to reduce the number of traffic accidents. Facial features retrieved by supervised gradient descent method Objectives Strengthening of technology transfer and commercialization capacities of partner HEIs. Transfer of existing computer vision applications from HEIs to SMEs in the automotive industry sector. Developing new traffic- and transport-related computer vision applications with commercial potential in collaboration with SMEs in the automotive industry sector. Headlight detection in urban and rural driving conditions. Methodology for testing and evaluation Research goals: definition of experiments which will be used for testing and evaluation of the developed driver assistance systems development of a system which uses only one video camera for vehicle detection and tracking on multiple road lanes; and computation of road traffic parameters (traffic flow, origin-destination matrices, vehicle classification, etc.) Moving object detection system Project coordinatorSenior researchersJunior researchers Prof. Sven LončarićProf. Slobodan RibarićProf. Ivan PetrovićProf. Hrvoje GoldNikola BanićDarko Jurić Administrative managerProf. Zoran KalafatićProf. Siniša ŠegvićProf. Niko JelušićJosip ĆesićKristian Kovačić Marijana Jurić FraculjProf. Edouard IvanjkoProf. Marko SubašićProf. Mato BaotićIvan FilkovićIvan Krešo Iva HarbašMarkan Lopar General information Project team Surround view parking assistance Developing methods for surround visualization of a vehicle using four digital video cameras mounted on the vehicle front, back and sides. The images are stiched together to form the top view of the vehicle. Vehicle surrounding coverage by cameras University of Zagreb, Faculty of Electrical Engineering and Computing Prof. Sven Lončarić University of Zagreb Faculty of Electrical Engineering and Computing Unska 3, Zagreb, Croatia Contact By using image and illumination statistical properties a significantly higher global illumination estimation accuracy has been achieved. By upgrading the Light Random Sprays Retinex algorithm a significantly faster local color and brightness adjustment has been achieved. Input image Resulting image Automatic headlight detection Automatic vehicle headlight detection and tracking is the base for the two following systems: automatic high-beam control and forward collision warning system. Detecting roadside vegetation can be used for increasing traffic safety (detecting vegetation that is occluding traffic lights or traffic signs), aiding autonomous vehicles in off- road navigation, and also for guiding service vehicles in roadside maintenance tasks. Detection of roadside vegetation Example of roadside vegetation detection Vehicle detectionVehicle tracking and counting