Automated Classification in Traffic Video at Intersections with Heavy Pedestrian and Bicycle Traffic
Sohail Zangenehpour, Luis Miranda-Moreno, Nicolas Saunier · PolyPublie (École Polytechnique de Montréal) · 2014
Pedestrians and cyclists are vulnerable road users and despite their limited presence in traffic events, these two groups have the most collisions resulting in injuries and fatalities. Due to problems regarding data collection for pedestrians and cyclists, there is a shortcoming in the field of road safety with regards to the availability and quality of data for non-motorized modes. Also, due to the constant change of orientation and appearance of pedestrians and cyclists, detecting and tracking them is a hard task. This is one of the reasons why automated data collection methods have mainly been developed to detect and track motorized traffic. This paper presents a methodology based on Histogram of Oriented Gradients to extract features of an image box containing the tracked object and Support Vector Machine as a classifier, to classify moving objects in crowded traffic scenes. This method classifies moving objects into three main types of road users: pedestrians, cyclists, and motor vehicles. This is done by first tracking each moving object in the video, classifying its appearance in each frame and then computing the probability of belonging to each class based on its appearance and speed. Bayes’ rule is used to fuse appearance and speed to predict the class for each object. Testing results show good performance, with an overall accuracy of more than 90%.