Edge Computing-Based Automated Vehicle Classification System Using the MobileNet V2 Model

Abstract

The volume of traffic in one day is referred to as the average daily traffic volume. The Average Daily Traffic System (LHR) is also used to detect road damage caused by excessive vehicle loads. In the LHR system, vehicle data is still collected manually, with humans calculating the type and number of vehicles based on observations made and then divided into a time span. As a result, a system with a camera and deep learning data processing is required to automatically calculate the type and number of vehicles. The goal of this research is to develop edge computing systems by improving the system's performance in the calculation and classification of vehicles using the SSD MobileNet V2 model. The results of the MobileNet model scenario 5 have the lowest loss value of the five scenarios. The MobileNet V2 model can better classify vehicle types with a 65 FPS inference process.