Papers by Carlos R C Souza
This paper presents a probabilistic logical model of traffic rules with the goal to provide highl... more This paper presents a probabilistic logical model of traffic rules with the goal to provide highlevel interpretation of the traffic rules regarding lane signalisation in video scenes, as observed from a vehicle’s viewpoint. The images are provided by a monocular camera attached to a vehicle driving in normal traffic situations. A low-level computer vision algorithm classifies the type of lane dividers and their relative positions with respect to the vehicle, then sends the information to be used as evidences by a probabilistic inference engine that reasons about the vehicle’s location and actions. The inference is accomplished using a first-order knowledge base formalising right-handed traffic rules. Uncertainties inherent to the sensors are treated within a probabilistic framework, the Markov Logic Networks, and results are compared to a well-known baseline classifier, the Naïve Bayes. We evaluate the use of these techniques under real-world traffic situations. Keywords— Spatial re...
Lecture Notes in Computer Science, 2011
This paper describes a probabilistic logic reasoning system for traffic scenes based on Markov lo... more This paper describes a probabilistic logic reasoning system for traffic scenes based on Markov logic network, whose goal is to provide a high-level interpretation of localisation and behaviour of a vehicle on the road. This information can be used by a lane assistant agent within driver assistance systems. This work adopted an egocentric viewpoint for the vision and the reasoning tasks of the vehicle and a qualitative approach to spatial representation. Results with real data indicate good performance compared to the common sense interpretation of traffic situations.
Uploads
Papers by Carlos R C Souza