A New Approach Algorithm for Counting of Vehicles Moving Based On Image Processing

• Published in: International journal of computer science and information security 2016-10, Vol.14 (10), p.366
• Main Authors: Setiyono, Budi, Ratna, Dwi S, Mukhlash, Imam, Augusta, Reza J
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Experiments; Methods; Vehicles
• ISSN: 1947-5500

The traffic problems are faced almost in every big city. This is due to the increase number of vehicles that potentially cause a traffic jam. Therefore, it is necessary to analyze the traffic density to obtain information, such as the number of passing vehicles, that information will become reference for taking some policies like traffic light time setting, road widening or the other policy. One of methods than can be implemented is by counting vehicle based on digital image processing, that more efficient than the manual counting. Research about image processing for counting has been done, but some of it depend on the video time taken. This research discussed about robust algorithm for counting vehicle in a traffic video for morning, afternoon and evening video. Processes such as background subtraction, noise removal, object detection and counting will be discussed therein. In the background subtraction method is used GMM (Gaussian Mixture Model), then the obtained result from GMM method will be processed in order to remove shadow that move similarly like a vehicle based on its intensity with frame mean and mode. While in object detection will be checked neighbors on each pixel of the image. Robust process was carried out by analyzing shadow that detected in GMM process, then shadow removal will be executed in order to obtain detecting and counting results that was accurate, this process called updating foreground process. Updating foreground process, will update the obtained result from GMM depend on mean and mode on present frame. The obtained result show that updating foreground result is more accurate than GMM.