Development of a New Tracking System Based on CMOS Vision Processor Hardware, Phase II Prototype Demonstration

Abstract

Intelligent transportation systems depend on being able to track vehicle operations and collect accurate traffic data. This project targets a hardware-based video detection system for real-time vehicle detection. To allow real-time detection, customized hardware implementation of the system is targeted instead on the traditional computer-based implementation of the system. The system includes four main processing steps. First, a camera is used to capture images. Second, the captured images are segmented using the Mixture-of-Gaussian algorithm. Without sacrificing the segmentation accuracy, researchers modified the Mixture-of-Gaussian algorithm to allow more efficient and economical hardware implementation in terms of design overhead and hardware resources. Third, the segmentation regions are extracted and validated as the objects of interests. In the last step, the validation result will be wirelessly transmitted to a variable message sign, which displays necessary traffic information. Since the system design includes integration of diverse devices, the video design kit from Xilinx is used. Such a hardware-based vehicle detection system has been experimented tested with practical videos of traffic scene.