Moving Vehicle Classification Using Pixel Quantity Based on Gaussian Mixture Models

Abstract

One problem of transportation that often happens is the traffic congestion. In order to address this problem, the information related to traffic are needed, such as type and total number of vehicles that passes certain road. This research discussed classification of the types of vehicles using pixel quantity. The Gaussian Mixture Models (GMM) used to extract foreground and background images. To classify vehicles, we use a quantity of pixels in which the amount is obtained based on the experiment. In the last stage, tracking and counting on vehicles passing through Region of Interest according to the classified type. The result is an algorithm capable for classifying type of vehicles with a high degree of accuracy. The experiments were carried out with two road conditions, namely a quiet and crowded road. On a quite road, the Kedung Cowek street and Wonokromo street, we obtained accuracy of 98.87% and 96.67% respectively. While on the crowded road, the Diponegoro street and Pemuda street, we get accuracy of 95.45% and 89.13%.