Video Traffic Parameter Detection System For Traffic Signal Control

The development of economy speeds up the progress of urbanism. As the rapidly increase of vehicles in urban, which caused a difficult traffic problem to all of the big city. Intersection is the city traffic node, the main location of flowing and the key section for congestion of the city. Therefore, the reasonable signal control plan of intersection has an important significance for traffic control. In order to provide reliable data for the selection and optimization of traffic signal control, we must identify the traffic condition timely and accurately. In reality, the number of installed traffic detectors is much less than the requirement of traffic signal control. On the other hand, there exists a considerable amount of video monitoring cameras in the network for purpose of traffic surveillance and monitoring. These cameras could provide a sequence of images of real-time traffic condition. If we can use these videos to estimate the real-time traffic condition, it is very meaningful for traffic signal control.In view of this, a new image processing method is proposed to estimate traffic flow parameters by using these videos and we use the detected parameters to optimize the traffic signal control plans of the intersection. The main contributions of this study are summarized as follows:(1) The cameras in road monitoring system are provided by different device providers, which makes these videos very difficult to be processed by using existing video-based methods, therefore we develop appropriate programs to obtain the available image data. The proposed method has practical advances as it could achieve accurate available traffic image data without any requirement of modification to existing video monitoring system.(2) In order to detect moving vehicle quickly, we present a samples-based adaptive segmentation detection algorithm for highway scene. Firstly, selecting the first frame as a referenced frame, and using a random policy to select values to build a samples-based estimation of the background. Secondly, adopting different strategies to update different regions, and using frames subtraction and background extend detection to classify the pixels in each frame into background area, uncovered background area and background extend area. Finally, using different rate to update these areas and virtual coil based detection algorithm to count the traffic flow.(3) We present a new traffic parameters extraction approach for traffic surveillance system at an urban intersection. Firstly, as the low occurrence probability of background pixels, presenting a selective background model (SBM) framework, then using the edge information and texture feature to select the appropriate candidate pictures and generate the background image. Secondly, segmenting moving objects by using background subtraction, then combining them with edge information and texture feature to determine whether there is a vehicle in the detection zone. On this basis, we use dynamic threshold to optimize the vehicle count result.(4) According to the spatial distribution of the video detector in existing road monitoring system, we present a new intelligent signal control algorithm to detect spatial and temporal correlation parameters. The upstream video detector detects the traffic parameters, and then notifies downstream signal controller which use the current traffic condition and the upcoming traffic flow to optimize traffic signal control plans.