Analysis of Classification and Calculation of Vehicle Type at APILL Intersection Using YOLO Method and Kalman Filter

Vehicles are a necessity that currently must be owned by the community, be it private vehicles or public transportation. However, with the increase in the number of vehicles, the government as the party that has responsibility for facilities and infrastructure does not keep pace with the increase in roads. This has resulted in frequent traffic jams, especially during busy times or working hours, thus overwhelming the authorized officers of both the Transportation agency and traffic police. To assist officers in the field in breaking down congestion that occurs with the help of CCTV monitoring installed in several sections of the APILL intersection, a real-time detection and calculation system is needed. The focus of the research is to optimize the Yolov3 algorithm in detecting and classifying vehicles by applying the Kalman Filter as a vehicle tracking method that has been classified. The dataset used for this research is the MSCOCO dataset with more than 300,000 object images. The test results of the two types of CCTV cameras that were carried out obtained the highest accuracy evaluation model of 53% when the APILL was red because the Kalman Filter did not work optimally.