Real-Time Vehicle Detection and Tracking System in Street Scenarios

The paper represents a framework for a vehicle detection, segmentation, and tracking system. The Challenge is to use a single monocular camera as input, in order to achieve a low cost final system that meets the requirements needed to undertake serial production in auto supervisory industry. We use a hierarchical method from the foreground region level to the vehicle level. The approach concerns stages of motion detection, edge detection, filtering, detection of the vehicle’s position, and investigation into tracking cars by their appearance visual features. Color, which is one of the strongest cues, is used for the tracking step. The Continuously Adaptive MeanShift Algorithm (CamShift) is an adaptation of the Mean Shift algorithm, and we use it as a tracking method. Competitive performance results are provided using real video sequences in real traffic conditions.