Implementation of Gaussian Mixture-Based Background Segmentation and the Level of Service of a Signalized Intersection

Abstract

The common major problem of all urban and developing cities or provinces is traffic congestion. Moreover, Surigao City, Surigao del Norte, Philippines is one of the developing provinces that experiencing congestion in particular Rizal St. and San Nicolas St. intersection which needs improvement. This study aimed to perform analysis of the intersection to rate its performance quality using Level of Service (LOS) based on the calculated control delay in seconds per vehicle. The Vehicle Analyzer Application (VAA) WAS created to show the number of vehicles detected by adjusting the threshold value from the slider. It WAS applied by Mixture of Gaussian with shadow detection as a background segmentation using the image processing techniques and algorithm (image segmentation, bilateral filter edge detection and contours) to be able to get the vehicle count on each lane and the classification of the vehicles. The SECOND video processed by the VAA classifier got the highest F1 score accuracy percentage of 80.86% taken last November 15, 2018 due to a fair weather, noise, small movements of the camera, occlusions, pedestrians passing IN the road, and the bit rate and resolution of the video that contributed error. Following the calculations of a Signalized Intersection Methodology of HCM 2000, the intersection has a control delay of 33.1 sec/veh and an LOS C which causes much higher delays that results noticeable congestion and fair progression of vehicles.