M-Eng in Electronic Systems 2013

Abstract — This paper presents a car recognition system by locating and segmenting tail lights in the night-time road environment. Numerous approaches towards car recognition during day time have been implemented in the past. However, the features of cars during daytime are seldom available when it’s dark due to lack of lighting and other conditions. Unlike previous work in the area, this system employs HSV color thresholding for segmenting the red regions (brake lights) and capturing certain object features of the segmented parts. These features are used to train and classify the different classes of lights for different car models. For this, the machine learning based approach – Support Vector Machines (SVMs) is incorporated. Satisfactory results were obtained and they show that the SVM technique is effective for classificatio