Modernising traffic flow analysis: A computer vision-driven prototype for vehicle detection

Abstract

RESEARCH QUESTION How can computer vision be used to support traffic flow analysis?

RESEARCH SUB-QUESTIONS What is required within the planning, development and implementation phases of a prototype that performs accurate vehicle detection? How will the performance of a vehicle-detection prototype be measured? What are the barriers to the development of a vehicle-detection prototype?

Computer vision holds the capabilities of performing duties by replicating tasks that the human visual system accomplishes. For this reason, computer vision is being employed as a tool to modernise and advance the management of traffic on a global scale. Traffic management remains an issue in many regions of the world, evidenced by barriers such as street obstacles, inefficient road signals, vehicles speeding, traffic congestion, and underdevelopment of freeways. Due to this, computer vision-driven management systems have been developed to combat such problems, demonstrated by their role in travel assistance and navigation, parking management and enforcement, real-time traffic control, and license plate recognition (Buch et al., 2011). Subsequently, this research explores how vehicle detection can be applied to support traffic flow analysis within Hawkes Bay, New Zealand through the use of computer vision approaches. This research assumes a design-based approach, comprised of two iteration-based approaches employed to develop a prototype for vehicle detection utilising an Nvidia Jetson Nano. The approaches are analysed according to accuracy, processing time, cost, and overall suitability. Results show that the prototype has great potential as an alternative approach to current traffic flow analysis. Finally, recommendations are offered for future research and other users working with similar devices.