Developing a Framework for Vehicle Detection, Tracking and Classification in Traffic Video Surveillance

Abstract

Listen

Translate article

Intelligent Transportation System and safety driver assistance systems are significant topics of research in the field of transport and traffic management. The most challenging of moving vehicle detection, tracking and classification is in the Intelligent Transportation System (ITS) and smart vehicles sector. This paper provides a method for vision based tracking and classify different classes of vehicle by controlling the video surveillance system. Several verification techniques were investigated based on matching templates and classifying images. This paper focused on improving the performance of a single camera vehicle detection, tracking and classification system and proposed a method based on a Histogram of Oriented Gradient (HOG) function that is one of the most discriminatory features to extract the object features and trained for classification and object on the Linear Support Vector Machine (SVM) classifier. Also categorize vehicles on the shape or dimension based feature extraction with cascade based Adaboost classifier which has the high predictive accuracy and the low cost of storage affirm the efficacy for real-time vehicle classification. In the final stage, for minimizing the number of missing vehicles, Kalman Filter was used to track the moving vehicles in video frame. Our proposed system is checked using different videos and provided the best output with appropriate processing time. The experimental results shows the efficiency of the algorithm.