A novel drive-through vehicle profiling system

Abstract

As commercially available off-the-shelf (COTS) vehicle profiling products have many limitations and restrictions, a better solution is expected from this industrial research project. This research aims to develop a novel integrated drive-through vehicle profiling system that automatically classifies incoming vehicles based on multiple factors by critical features of different car type. This system should be able to accurately measure the length and width of moving vehicles and to recognize some critical features to be assigned to appropriate size categories. Many possible solutions have been proposed in the past decades. The main technological approaches for detection can be divided into two: sensor-based detection methods and vision-based detection methods. Sensor-based detection methods collect different types of data to perform tasks. Vision-based detection methods can solve complex tasks, such as face detection, traffic sign detection and pedestrian detection, etc. With low price tag and easy installation, a vision-based sensor is a natural solution for detection. The tools for this research are based on the latest technologies, such as Light Detection and Ranging (LiDAR), Computer Vision (CV), geomagnetic sensor, and the deep learning technique. In order to cater for the needs of Smart City development, they can also provide all the profiling data that can be shared with other systems to form big data. This research will develop the theory and methodology for a new vehicle profiling system design with lower power consumption, more flexibility, and higher cost-effectiveness. The new system is expected to be easier to install and to generate significant savings on maintenance and total cost of ownership.