IMPLEMENTATION OF A VIDEO ANALYTICS-BASED VEHICLE TRACKING SYSTEM FOR COUNTING VEHICLES BASED ON MOVEMENT DIRECTION AT A FOUR-WAY INTERSECTION

• Main Author: Seipanya, Ayub
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/85274
• Institution: Institut Teknologi Bandung

Congestion is an important issue that has various adverse impacts on society and needs attention. Better traffic management efforts can be undertaken to reduce congestion by paying closer attention to the traffic conditions of a traffic space relative to neighboring traffic, especially at intersections. Observations of these conditions can begin by identifying traffic parameters, particularly traffic volume, and it is necessary to know the number of vehicles passing through a traffic space to obtain the traffic volume value. Counting the number of vehicles in traffic is usually done manually, but with the advancement of artificial intelligence technology, vehicle counting can be carried out using object detection technology. Moreover, object detection technology combined with object tracking technology can make vehicle counting easier and more accurate. This study discusses the development of a vehicle tracking system to count vehicles at intersections by applying the V-Model waterfall methodology, which consists of communication, planning, modeling, construction, and evaluation. The system implementation is carried out according to the requirements and designs that have been established, with the implementation stages consisting of dataset management, model training, and vehicle tracking development. The test results of this system show good detection, classification, tracking, and vehicle counting capabilities, marked by high evaluation metric values, both for the system using ByteTrack and the system using DeepSORT. However, the system using ByteTrack has higher evaluation metric values compared to the system using DeepSORT. Nevertheless, the system does not work optimally when dealing with data contexts containing occlusions, such as between vehicles and light poles.