THE PUBLISHING HOUSE PROCEEDINGS OF THE ROMANIAN ACADEMY, Series A, OF THE ROMANIAN ACADEMY Volume 15, Number 1/2014, pp. 76–84 INTELLIGENT SURVEILLANCE SYSTEMS FOR DETECTION AND MULTI TRACKING OF VEHICLES IN REAL TIME Saeid FAZLI 1 , Morteza RAHMANI 2 , Shahram MOHAMMADI 1 1 Zanjan University, Electrical Engineering Department, Zanjan, IRAN 2 Payame Noor University, Department of Electrical Engineering, PO Box 19395-3697, Tehran, Iran Corresponding author: Morteza RAHMANI, E-mail: Rahmanim@znu.ac.ir Given the improvement of digital , s cameras and the system of video , s processing, it have been considering the civil controlling and managing and observation into the traffic. Among this, the computer vision systems have used for creating of some equipments of automatic controlling of traffic. In this article, we have presented the new algorithms by techniques of image processing for an intelligent system of transportation. It can be done some operations of observation into traffic in the real time and by using of a kind of camera, small angle and with the least obtaining information of scene. The applied technique in this study is based on the tracking of optical flow by Horn & Schunck. In order to do this, we selected a part of highway and then obtained the speed for every pixel by using of calculation of derivation of image into the place and time. At last, the system of transit vehicle detected and tracked in the real time and specified the number of those vehicles. Key words: camera, image processing, optical flow, vehicle, track. 1. INTRODUCTION The growth of going bade and forth in the highways and civil roads have been explosively recently. Even the manual and traditional systems of traffic controlling aren , t responsible to this situation at all. The above factors and a lot of other parameters have led to use of intelligent systems of transportation which is a kind of interdisciplinary technology in the analysis and inspecting of system of traffic [1, 2, 3, 4]. Today, by improving of systems of video , s processing, the intelligent method has been increasing for detection and tracking of motion objects. Among this, because of improvement of systems of transportation and getting to parameters of traffic, those intelligent systems of transportation have been considering very much. Observational eco-systems are done by using of any camera that is one of the most applicable device in two recent decade [2]. In a system of traffic, obtaining of parameters such as estimation of movement, speed and sequential distance between vehicles have a key roles. One of the important work in the analysis and video , s observation is estimating of movement. For detecting of object , s movement, calculating of pathway and speed of object, it should be using of founding field of movement. One of the best method for perceiving of surrounding and those features of intelligent system is two-dimensional images. In the systems based on machine vision, obtaining information and processing of available parameters in the image are among the important factors for observation. Adverting of some high-speed processors allow to process sequencing of images. Since the speed, movement and position of objects is receptive by using of video move, so certainly the available information in sequencing of images are more than those in a single image. One of the most essential methods of extracting information refers to sequencing of images is assigning the speed and sort of movement of current objects in scene of image-tacking. In order to do this, There has been presented some several algorithms by different people. One of the available methods is that offered by Gupte et al. [5]. In this method, the whole part of vehicle have displayed as a single point. But this algorithm has a problem for considering of shadows, occlusions and large vehicles. Another method are contour tracking which has presented by Koller et al. [6]. In this presented model, at first one of the images is setting and selecting from background and then the operation of tracking are doing by detecting the border of objects. However, the main problem of this method refers to occlusion also the algorithm of contour are so sensitive to tracking frames.