A Qualitative Traffic Sensor based on Three-Dimensional Qualitative Modeling of Visual Textures of Traffic Behavior
E. Bonet, Silke Fresno Moreno, F. Toledo, G. Martín · Industrial and Engineering Applications of Artificial Intelligence and Expert Systems · 2022
This paper introduces a camera-based traffic sensor which uses qualitative image processing to provide a qualitative description of the state of traffic. The video input is processed by means of a temporal Gabor transform, and qualitative traffic textures are recognized in the temporal Gabor space. The resulting qualitative traffic textures are matched with a qualitative model of urban traffic behavior in video sequences, in order to obtain the qualitative description of traffic behavior which better explains the observed textures. The matching process is performed focusing only in discriminant regions of the video space, which are obtained from the envision graph of the qualitative model, resulting in a great gain in performance. The proposed method obtains an exceptionally good response to low or null movement congestive situations, which most state-of-the-art camera traffic sensors based on movement analysis fails to perform.