Study On Method For Traffic Intersection Vehicles Detection Based On Single Camera

Following the large expansion of our cities, the urban traffic problem has been moreand more serious. So, ITS has been introduced in urban traffic control and management,with the target that, human can enjoy the traffic in their city life. The base of ITS include thecollection of traffic basic data with high accuracy and timeliness. Recently, the vision baseddetection technology of data collection for traffic, has become a development direct oftraffic data collection, and has been seen as the hot issue in computer vision and ITSresearch area.Traffic insertion is the key of whole urban traffic. Therefore, how to apply visiondetection technology into the data collection in traffic insertion has becomes a new hot topic.On the basis of existing detecting method of traffic insertion, a innovative solution has beenformulated, which can track vehicles and collect direct information in traffic insertion fromone arterial road approach by using single camera. As a result, ITS can calculate minimumgreen time or select key flow for the traffic insertion. The new solution is of veryimportance of realism, because it will reduce the cost of collecting data process obviously.The paper analyses several implementing schemes of single camera detection in trafficinsertion, and finally choose the most feasibility one to carry on further investigation. To thebusy and complicated traffic insertion, an area based method background update is proposed,which is based on mixture Gaussian background model and draws support from the estimateof traffic lights. And, the paper proves the validity of the tracking algorithm throughexperiment, which can track vehicles by using Kalman filter theory to predict position andhistogram feature-matched method to match vehicle target.