CIAM-COR-R33

Research Team

Fei Dai, West Virginia University

Funding Sources

West Virginia University Core Funds — $128,480

West Virginia University Match — $128,480

Total Project Cost — $256,960

Agency ID or Contract Number

69A3551847103

Start and End Dates

10/01/2021 — 09/30/2023

Project Description

The objective of this project is to develop a pragmatic oversize vehicle warning system for avoidance of low clearance collision based on the PI’s prior work and state-of-the-art deep learning and single view geometry in computer vision.

In the proposed system, a surveillance camera is installed on an overpass or mounted on a pole along the road in such a way that oversize vehicles are detected early and provided with enough time to react. The system requires one camera per bound, with a fixed view of all lanes. Once a camera is set up, the Manhattan world (i.e., all surfaces in the world are aligned with three dominant directions, corresponding to the X, Y, and Z axes) in the surveilled traffic scene is constructed for the purpose of transforming coordinates from an image plane to real world planes as well as converting measurements from pixel units to metric units. Then, incoming camera frames are processed to detect and track vehicles. Detected vehicle regions are further processed to discover their heights, widths and lanes. The location of the vehicle is also calculated based on the constructed Manhattan world to identify its lane. Oversize vehicles are classified simply by comparing the estimated dimensions with the overpass clearance or tunnel entrance dimensions. Once an oversize vehicle is detected, the warning system will be activated. The specific objectives of the project are:

  1. Validate existing algorithms and develop new ones to simultaneously detect and track vehicles in real-time and locate them on a road-plane coordinate.
  2. Develop an algorithm to extract height and width of a vehicle from video frames and investigate on appropriate positions and view angles of a camera to obtain the most accurate measurements.
  3. Develop and provide a warning system and network for its communication with cameras and verify their effectiveness.