Research On Video-based Vehicle Detection In ITS

Intelligent Transportation System (ITS) is a new focus of research and application in the field of transportation. It's the integration of much advanced multi-disciplinary technology, such as Computer Technology, Image Processing, and Artificial Intelligence. This system has obvious advantages in information processing and great potential for future development. As an essential part of ITS, Video-based vehicle detection method has unparalleled advantages over the traditional vehicle detection methods. However, as Video-based vehicle detection method is subjected to the control of environment, climate and other factors, the technology needs to be improved continuously.Video-based Vehicle Detection method is composed of Image Acquisition, image pre-processing and image analysis (extraction of the target vehicle) in general. In this paper, a lot of research and exploration are done on the key issues about the three-part of the paper. A new feasible method is proposed based on much theory and practice, the feasibility of which is proved by experiments then.The main contents of this article are as follows:First of all, the image is divided into two categories in light of the actual traffic images: the foreground image and the background image .when the static background is removed by the classical background difference-based method, we can get the moving foreground. During this process, we use the probability statistics-based method of statistical averaging to construct the model for the background and a new Pixel-based classification method is proposed to update the background. We can prove that the background is more accurate and the phenomenon of "fusion" can be reduced effectively when compared to the traditional methods via experiments. Second, aiming at the co-existence of shadow and the vehicle in the foreground image, the paper proposes a new method to remove the shadow basing on the principle of optics. In this method, we use the brightness of the corresponding pixels of the moving target and the background which belongs to the current frame that has been differentiated by background difference-based method to make the judgment. We can extract the vehicle by partitioning the vehicle and shadow in accordance with the difference on reflectance and illumination. The large number of pictures that used for image processing and analysis in the research are taken from the true image of the traffic by the author. Experiments prove that the new method is effective and can achieve the desired requirements; the practicality and strong robustness are achieved as well.