DL-based commercial vehicle characteristics identification, detection, and classification system

Information is everywhere, especially in the commercial vehicle industry. Vehicles may be classified by number of axles/tires. There are several text- and label-based classification systems: for dangerous goods transport (HAZMAT); vehicle safety code compliance (CVSA); and general identification and tracking (license plates, USDOT numbers). Employing humans to perform simple classification and recognition tasks can be impractical. However, explicitly programming these tasks can be challenging. We propose research into object/character detection and recognition methods to develop a fast, accurate, and robust identification system for all aforementioned vehicle characteristics. We will utilize deep learning, in which machines learn patterns from data inputs and desired outputs. IRD will provide datasets from which we can develop our models, and will benefit by adding our solution to their many Intelligent Transportation Systems. We foresee that our system, allowing fast commercial vehicle identification and tracking, will be deployed throughout Canada, North America, and the driving world.

Faculty Supervisor:

Seok-bum Ko

Student:

Juan Yepez;Riel Castro-Zunti;Hao Zhang

Partner:

International Road Dynamics Inc.

Discipline:

Engineering - computer / electrical

Sector:

Automotive and transportation

University:

University of Saskatchewan