Classified Counting and Tracking of Local Vehicles in Manila Using Computer Vision
Febus Reidj G. Cruz, Carissa Jane R. Santos, Larry A. Vea · 2019
Many countries have improved their traffic surveillance system by using computer vision to classify and track different types of vehicles. Having this data can lessen management cost and help improve rules and regulations in route planning. In Manila, using the common vehicle type dataset for traffic management is inefficient. The city has at least 9 types of vehicles present in its main roads, and at least 21 types of vehicles in secondary and tertiary roads. By using digital image processing, an algorithm for classified counting and tracking was created. The algorithm utilizes machine learning methods to create a local dataset with 16 types of vehicles for the city. After successfully creating the dataset, the system can detect all the present vehicles on the selected footage accurately. In a recorded video containing 11 types of vehicles, 92.96% were correctly classified and 95% were counted. The location of the Region of Interest (ROI) for counting must be strategically placed to avoid misclassified counting. The local dataset can also be improved by collecting data from other roads, and by adding the other local classes such as tricycles and pedicabs.