Traffic evaluation using camera recording

Abstract

This thesis explores the potential of using camera recordings and artificial intelligence for vehicle data acquisition, comparing modern AI-based methods with traditional measurement techniques. The research combines basic principles of image capture with machine learning applications, particularly implementing YOLOv12 for traffic video analysis. Our approach evaluates both conventional measuring technology (optical speed sensors, triaxial accelerometers, steering sensors) and camera-based detection methods using recordings from road intersections in Pardubice. While the methodology framework was successfully established, including data preprocessing, event detection, annotation, and model selection, the data acquisition process could not be completed due to restrictions imposed by the Pardubice city police changing the permitted period for data access and other legislative complications related to privacy regulations. Despite these challenges, the study provides valuable insights into the capabilities and limitations of AI-based traffic monitoring systems compared to conventional measurement methods.