Video-Based Vehicle Signature Analysis and Tracking System Phase 2: Algorithm Development and Preliminary Testing

Problem

This report describes the field tests of the V2SAT video based vehicle signature analysis and tracking system. This system employs video cameras positioned over roadways to obtain images of passing vehicles. The video data is then processed by a digital image analysis to develop a "Video Signature Vector," or VSV. The VSV consists of information such as color of the vehicle, size of the vehicle, and other characteristics like general body shape information.

Using multiple arrays of video cameras placed at various points along a freeway connected to a central computer via a wireless network, VSVs can be compared from the video cameras enabling individual vehicles to be tracked through the instrumented region of freeway. The data can be used to validate traffic flow, generate origin-destination data, validate local modal emission models, and possibly be implemented for law enforcement purposes.

Phase 1 of this research focused on validating the operational concept of the proposed V2SAT method. This included testing the viability of the VSV concept.

The primary focus of this part of the research (Phase 2) was to develop and debug the platform for the field implementation of V2SAT. This platform included software for real-time image processing and storage, refinement of the image analysis algorithms, and the required hardware. The system was then field tested under actual freeway conditions. The test site was a pair of freeway overcrossings .34 miles apart. The system was connected using a wireless network.

Findings

It was determined that the system had the ability to correctly re-identify vehicles at successive sites was 93.6%. Difficulty was encountered with motorcycles and tandem trucks