On-Road Vehicle Detection using Support Vector Machine and Decision Tree Classifications

On-road vehicle detection is a major part of various applications, such as driver assistance systems and auto-driving, and these systems need to detect vehicles robustly and accurately. This paper proposes a robust vehicle detection system to detect the front vehicles by using a single camera mounted on the car. The proposed system consists of two main steps, which are, hypotheses generation (HG) and hypotheses verification (HV). The first step is to find the candidate regions to the vehicles in the image. The guide to these regions is the shadow underneath the vehicle because it is always darker than the road surface. The system fits the generated regions with the width of the vehicle and reduces the number of hypotheses by calculating the entropy in two different ways. The second step is to verify whether the generated hypotheses contain a vehicle or not, and this done by Histogram of Oriented Gradients (HOG) to extract the features. In designing the proposed system, the Support Vector Machine (SVM) and Decision Tree (DT) classifiers are used for classification. Experiments were conducted using the challenging GTI DATA database to ascertain the usefulness of the approaches. The methodology was evaluated against the state-of the-art and it was found that the proposed approaches produce outstanding results.