International Journal of Computer Applications (0975 – 8887) Volume 104 – No.6, October 2014 10 Determining Microscopic Traffic Variables using Video Image Processing Abdulrazzaq Alkherret Graduate Ph.D Student Faculty of Engineering Cairo University, Egypt Al-Sayed A. Al-Sobky Assistant Professor Faculty of Engineering Ain Shams Univ., Egypt Ragab M. Mousa Professor Faculty of Engineering Cairo University, Egypt ABSTRACT Vehicle detection and tracking play an important role in traffic management and control. Among available techniques, Video Image Processing (VIP) is considered superior due to ease in installation, maintenance, upgrade, and visualizing results while processing recorded videos. In this paper, a multiple-vehicle surveillance model was developed, using Matlab programming language, for detecting and tracking moving vehicles as well as collecting traffic data such as traffic count, speed, and headways. The developed model was validated for different lengths of region of interest (ROI), ranging between 5 and 30 m. Validation was established using simulated video clips, designed in VISSIM, and traffic data obtained from model were compared with actual measurements reported by VISSIM. Vehicle counts (or detections) obtained from the model are identical to actual counts. Comparison of speeds confirmed the model validity, especially with 10 m and 15 m ROI lengths. For these lengths, the mean difference of speeds is not significant at 5% significance level. Validation headway measurements was also confirmed for ROI of 10 and 15 m. With such successful validation, the model features many applications. Beside traffic data collection, the model can be applied for incident detection, speed enforcement, intelligent transportation system, etc. However, the model was validated assuming no lane changes. Camera position was also set to avoid overlap of vehicles. Accordingly, the model validity is limited to these assumptions. Further research is currently in progress to extend model validity to lane changes and different camera positions. Keywords Matlab, Image Processing, Traffic Surveillance, Vehicle Detection, Vehicle Tracking, Speed, Headway. 1. INTRODUCTION The literature is abundant with researches that dealt with detection and tracking of moving objects in video sequence, and numerous mathematical models have resulted out of these studies. A general subdivision of object detection techniques can be made of three main categories, namely, (a) Background Subtraction, (b) Temporal Differencing, and (c) Optical Flow. Similarly, the object tracking can be divided into three main categories, Point tracking, Kernel tracking, and Silhouette tracking [1]. In object detection techniques, Background Modeling (Background Subtraction) is used to detect moving objects in an image by taking the variations between the current image and the reference background image in a pixel-by-pixel fashion. The background subtraction method uses a simple algorithm. However, it is very sensitive to changes in the external environment. Similarly, Temporal Differencing method is used to calculate the absolute differences between two consecutive images to extract moving regions and obtain a threshold function to determine changes. The temporal differencing has a strong adaptability for a variety of dynamic environments, but its method of calculation is generally difficult to achieve complete outline of moving object. The Optical Flow method uses the optical flow distribution characteristics of moving objects over time in an image sequence. Flow computation methods cannot be applied to video streams in real time because they are very complex and very sensitive to noise [2] and [3]. In object tracking manners, the Point tracking method is based on monitoring and comparing the positions of different detected points from one frame to another. Kernel tracking method tracks objects by calculating the motion of an object shape and its appearance in successive frames. The Silhouette tracking method uses information inside the silhouette's region in the form of edged maps to track the object using shape matching approach [4] and [5]. As mentioned above, background subtraction methods are very sensitive to changes in the scene. Also, this method requires a training period absent of foreground objects and too slow to be practical. Stauffer and Grimson [6] modeled each pixel in an image sequence as a mixture of Gaussians and used an on-line approximation to update the model. Then, the Gaussian distributions of the adaptive mixture model are evaluated to determine which are most likely to result from a background process. Finally, each pixel is classified based on whether the Gaussian distribution which represents it most effectively is considered part of the background model. Kaewtrakulpong and Bowden [7] improved the previous adaptive background mixture model. The authors updated equations and utilized different equations at different phases to make their system learn faster and more accurately as well as adapt effectively to changing environments. However, Matlab [8] adopted the previous two studies and presented a System object to detect foreground using Gaussian Mixture Models (GMMs). Nowadays, detection and tracking moving objects is becoming more essential to traffic engineers since available systems such as video image processing (VIP) are successfully used in traffic data collection and traffic surveillance. According to [9], [10] and [11], all detector technologies and particular devices have limitations, specializations, and individual capabilities. Among these technologies, only microwave radar, active infrared, and VIP