IMPLEMENTATION OF TRAFFIC CONGESTION CLASSIFICATION METHOD FROM CCTV VIDEO BASED ON IMAGE FEATURE ANALYSIS WITH YOLO ALGORITHM

• Main Author: Fernaldy
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82090
• Institution: Institut Teknologi Bandung

Traffic congestion is one of the major problems in the field of transportation as it brings numerous negative impacts. With the help of artificial intelligence, classifying traffic congestion in the traffic CCTV video is possible to be done. When congestion is detected, the authority can then be notified, so that the occurring traffic congestion can be handled immediately. The YOLO algorithm can be used to segment road areas and detect vehicles captured in CCTV video. The traffic features can be measured based on the result of vehicle detection process to obtain a thorough understanding of the traffic condition. There are four traffic features, that is, traffic flow, occupancy, density, and speed. The traffic features can then be used to classify the traffic status. The Python programming language is utilized to implement the traffic congestion classification method. The road segmentation and vehicle detection models are obtained from the YOLOv8 pretrained model training processes. The traffic flow is measured from the number of vehicles detected, the traffic occupancy is measured from the ratio of the vehicle pixels to the road area pixels, the traffic density is measured by extracting the reciprocal of correlation property from the gray level co-occurrence matrix (GLCM), and the traffic speed is measured by utilizing pyramidal Lucas-Kanade optical flow. The measurement results can then be used to classify the traffic status by utilizing artificial neural network. Based on experiment, the traffic congestion classification method using the YOLO algorithm for road segmentation and vehicle detection can classify the traffic status with an accuracy of 84.75%, precision of 84.66%, recall 84.75%, and F1-score of 84.69%.