Automatic Localization Of Traffic Signs In Panoramic Vision

With the prosperity of the automobile industry, the improvement of people’sliving standards, and the gradual improvement of per capita vehicle ownership rate,traffic volume constantly increase, which leads to the t traffic accidents. If we areable to detect and locate the traffic signs timely during driving and to give feedbackinformation of traffic signs to the driver, driving tips and warnings will be given intime which can reduce the traffic accidents. So real time location of traffic signs isvery important. The key of location is automatic detection and tracking of trafficsigns. Panoramic vision simulates the scene information of360degrees aroundviewer which is much larger than the information of one single image. We do notneed to concern about traffic signs, which are out of the image when we collectthem through the panoramic vision. Accordingly, studying automatic locationmethod of traffic signs based on the panoramic vision in this dissertation ismeaningful.The contents of this paper are listed as follows:1) Analyze several common detection methods of traffic signs based on the color,shape and other characteristics of road traffic signs. Since Haar features areeasy and fast, we apply them to the detection and classification of traffic signsand use these features to train the detection classifier.2) Improve the traditional Mean-Shift tracking algorithm. Combine momentinvariants with the algorithm to realize tracking of traffic signs.3) Designe a solution to object tracking based on the relationship among thecameras in panoramic vision, to achieve automatic and continuous tracking ofthe object among multiple cameras using the improved Mean-Shift trackingalgorithm. Then according to the position of the traffic signs in images, the fastlocation of traffic signs in panoramic vision is carried out.