Research And Implement Of Highway Violation Identification Technology Based On Vision Drone

This article uses visual drones to patrol the airspace according to a certain route.It recognizes the two scenarios of vehicle rolling solid line and illegal solid line,and detects the license plate number of the illegal vehicle.Compared with the traditional illegal recognition technology in the transportation system It has a better effect on the accuracy and real-time identification of illegal rolling lines and crossing lines,and provides an efficient and flexible monitoring mode for traffic monitoring and management.First of all,for vehicle rolling line recognition,target detection for moving vehicles,the classic target detection algorithm cannot be applied to the scene of unmanned aerial vehicle road surface changes,so the selection based on the YOLOv3(You Only Look Once verson 3)detection algorithm,Collect vehicle data through drones,and use the graphics card for training to obtain the vehicle detection model.Then use the traditional image processing method combined with the adjacent angle averaging algorithm to get the lane solid line.Afterwards,a mathematical model of rolling line is constructed based on the diagonal of the vehicle detection frame and the solid line of the lane to determine whether the vehicle is rolling.The experimental results show that the vehicle rolling line recognition algorithm used in this paper has higher recognition accuracy than the traditional vehicle rolling line recognition algorithm.Secondly,according to the vehicle rolling line recognition result,determine whether the vehicle violates the violation of the solid line.The vehicle rolling line triggers the Deep-Sort(Simple Online and Realtime Tracking with a Deep association metric)algorithm to track the rolling line vehicle.The vehicles during rolling line are matched by Scale-Invariant Feature Transform(SIFT).The vehicles with high matching degree are given the same ID,and then the 20 th frame image before and after the rolling line of the same vehicle is compared.The vector direction connected between the center and the rolled lane line determines whether the vehicle crosses the line in violation of regulations.The experimental results show that,compared with the traditional cross-line recognition method,this method has a higher accuracy.Finally,in order to identify the illegal vehicle license plate,first locate the precise position of the license plate,and then correct the license plate under the view angle of the drone through the license plate tilt correction method based on the texture direction field.Gate Recurrent Unit(GRU)recognizes license plate characters.The experimental results show that the accuracy of the license plate recognition of the UAV is high in the scenario where the flying height is less than 8 meters and the flying horizon angle is less than 60 degrees.