Research On Vehicle Traffic Based On Video

With the development of the city. The contradiction between the vehicle and road is increasingly prominent. So,it is a burning question that how to make full use of existing resources to achieve Intelligent Transportation Systems. Intelligent Traffic Signal Control System as an important part of Intelligent Transportation System play an irreplaceable role in reducing the overall transit time of vehicles at intersections, improving traffic efficiency and reducing fuel consumption and environmental pollution,etc. Based on video vehicle detection and calculation problem is an important foundation for visual foundation such as Intelligent Traffic Light Control System, and have an important research significance and practical value.Currently, the measurement and calculation of the vehicle in intersection used mainly road contact physical detection devices to detect and calculate, which are based on pressure, electric or magnetic fields sensing devices. However, the installation and maintain of this kind of devices is difficult to achieve and the devices is easy to age so as to raising maintenance costs. But this kind of method does not solve the problem of traffic diversion. With the development of video detection technology, the calculation method of based on video detection and calculation as one of the method of non-contact vehicle detection,which is simple to install and maintain. So, the maintenance cost of the devices is lower. And this method can solve the problem of multilane and vehicles diverted. So it will gradually replace traditional vehicles detection methods become the focus of future research. However, video-based vehicle detection technology is still faced with many of hard technical problems such as dynamic background interference, static vehicle target detection, target occlusion shadow suppression,etc.To solve the above technical problems, this thesis proposes to utilize the road split to solve the interference of dynamic scenes of non-background,such as Branches sway. Utilize the algorithm of based on the color characteristics of the vehicle detection to detect static vehicle targets. By take the shaded area of the road into the road background model, realizing shadow suppression. And make use of color constancy algorithms to remove the interference of color characteristics of the vehicle and the road surface caused by the gradient of light source color. Finally, we improved the model of the lane space occupancy ratio to measure the lane of traffic density. Specifically, the main works and innovations of this thesis are as follows: l.An in-depth analysis of various classical edge detection operators and Hough transform line detection principle, and utilize the Hough transform to segment the ROI of pavement so as to eliminate the interference of dynamic scenes of non-background.2.An in-depth analysis of the color distribution of the vehicle and the road surface in RGB, and transform RGB into a new colour feature space, which the color characteristics of the road surface distribution is more compact. In this color feature space, we designed a rules to classified the road and vehicle pixels based on Bayesian classifiers, then convert the input image to a binary image annotation. Finally, the use of the image segmentation of based on energy minimization graph cut segment the vehicles.3.For the change of the color characteristics of the vehicle and the road color caused by the gradients of the light source color.A new color constancy is proposed to solve the multisource and non-uniform illumination scenes. First, partition the image into multiple smaller patch:then, traditional color constancy are applied on every patch to estimate the local illuminant; last, combining the local illuminant into a composite color of the illuminant by weighting their contribution and treating the illuminant as the best approximate illuminants of the scene to reproduce the image.This method is a effective way which breaks out the restrictions of the spectral power distribution of the light source and the reflective properties of the surface of object.Finally, the color constancy algorithm is applied to vehicle detection algorithm for testing whether the experiments of vehicle detection be improved.4.After an in-depth analysis of the model of lane space occupancy ratio and its existing problems, we proposed a improved lane space occupancy ratio model to measure the lane of traffic density by blending the space-time relationship of image capture device and the object.