Proceedings of The 3rd International Conference on Innovation in Computer Science and Artificial Intelligence
Intelligent Transport System for the Prevention of Road and Highway Accidents
Bencheriet Chemesse Ennehar, Boukemoum Zakaria and Menai Mohammed, Belhaddad Samir
Intelligent transportation systems become in recent years a research field of primary importance for the scientific community and for public authorities this interest mainly due to the number of deaths caused each year by road accidents. The main objective of this work consists of detection and tracking vehicles (mainly cars) in road and highway using HOG features and kalman filter. At first, Histogram of Oriented Gradient (HOG) is used to extract feature vectors. Thereafter, a dataset composed by HOG feature vectors of positive and negative examples will be used in SVM (Wide Margin Separators) training. The model obtained after training will allow us to classify the objects detected in vehicles / non-vehicles. For the tracking of detected vehicles, a motion model based on adaptive kalman filter is established. Experiments show that the combination of HOG features and Kalman filter considered as collaboration of a good feature descriptor and a good motion predictor gave good results regarding the cars detection and tracking also the prediction of Kalman filter model provides a reliable region for eliminating the interference of shadows and sharply decreasing the false detection rate.
Keywords: Vehicle detection, Vehicle tracking, HOG features, SVM classifier, Kalman filter.