#TITLE_ALTERNATIVE#

• Main Author: HAMONANGAN L. TOBING (NIM: 13206092) Pembimbing : Dr. Ir. Hendrawan, PETER
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/16818
• Institution: Institut Teknologi Bandung

Road traffic monitoring and analysis system main functions are to collect <br /> <br /> <br /> <br /> <br /> measurement data and to provide statistical report on road utilization. It is very <br /> <br /> <br /> <br /> <br /> useful in cities with dense population and busy land-vehicle activities. Such <br /> <br /> <br /> <br /> <br /> system typically consists of a sensor for detecting vehicles on the road and a <br /> <br /> <br /> <br /> <br /> database system for storing and retrieval of information. <br /> <br /> <br /> <br /> <br /> Among the sensors capable of detecting vehicles on the road, camera is the most <br /> <br /> <br /> <br /> <br /> popular device these days. This is due to the competitive price and the easy <br /> <br /> <br /> <br /> <br /> installation of the camera compared to other devices performing practically the <br /> <br /> <br /> <br /> <br /> same function. Image processing, especially object detection and tracking, is <br /> <br /> <br /> <br /> <br /> therefore the key procedure involved in developing such system. <br /> <br /> <br /> <br /> <br /> This research focuses on the design of a software prototype that can collect road <br /> <br /> <br /> <br /> <br /> traffic statistic as well as broadcasting real time road traffic jam status on the <br /> <br /> <br /> <br /> <br /> internet. This research uses background subtraction method for detecting cars in <br /> <br /> <br /> <br /> <br /> road lane at day time. For detecting cars at night, new method is proposed using <br /> <br /> <br /> <br /> <br /> the feature of car’s front lamps. <br /> <br /> <br /> <br /> <br /> The research also shows that although background subtraction method works well <br /> <br /> <br /> <br /> <br /> to detect cars at day time when they are separated, it doesn’t give the same <br /> <br /> <br /> <br /> <br /> performance when the method is applied to overlapping or joined cars seen by the <br /> <br /> <br /> <br /> <br /> camera. The car detection accuracy rate at day time is 86,7% when road traffic is <br /> <br /> <br /> <br /> <br /> at low density, but reduced to 64,2% when road traffic is dense. Meanwhile, the <br /> <br /> <br /> <br /> <br /> car detection at night has better performance at detecting cars, with high detection <br /> <br /> <br /> <br /> <br /> accuracy rate at 96,3%. The average processing rate is 0,061 seconds/frame for <br /> <br /> <br /> <br /> <br /> background subtraction method, and 0,021 seconds/frame for the newly proposed <br /> <br /> <br /> <br /> <br /> method for detecting cars at night. The testing is based on traffic videos taken at <br /> <br /> <br /> <br /> <br /> Jl. Thamrin, Jakarta Pusat. <br /> <br /> <br /> <br /> <br /> The software is built in C++, using OOP concept, on Windows Platform using <br /> <br /> <br /> <br /> <br /> MFC library, OpenCV library for image processing purpose, MySQL for storing <br /> <br /> <br /> <br /> <br /> and retrieving database, and HTTP to broadcast real time jam status on Twitter.