Research On License Plate Recognition Based On UAV Video

With the rapid development of China's economy,the number of cars of people holding are also showing growth.Because the number of cars increase rapidly,transportation network get more and more complicated.At the same time,security risks more and more serious.It's an urgent problem to be solved that how to get traffic monitoring more flexible and effective.By using UAV's lightweight,flexible and other advantages,combination of UAV technology and traffic management system,can overcome the shortcomings of traditional human patrol and fixed surveillance cameras,provide a new way for traffic management,improve and guarantee the effectiveness of the monitoring system.The license plate recognition has always been the research hotspot in the field of intelligent transportation.With the unmanned aerial vehicle technology in the field of intelligent transportation,the combination of the two is inevitable research trend.This paper designs a license plate recognition system based on UAV video.When processing the traffic video collected by UAV,this paper position vehicles first,and then identify the license plate.By using this way,car detection and license plate recognition can be done easily.So this system is divided into two parts,namely vehicle detection and license plate recognition.Vehicle detection completes the function of car positioning.Because the difference between video collected by fixed camera and video collected by UAV is that background is dynamic.Object detection based on static background can't be used,and at the same time,object detection based on dynamic background can't detect the static car in the video,so this paper uses the extracted image HOG characteristics and SVM to detect cars.This paper uses the most popular method to recognize cars' license plate.Firstly,the license plate location is combined with the HSV color feature and SOBEL edge detection method.The tilting license plate is calibrated by Radon transform and so on.Then,the license plate character segmentation is carried out by vertical projection method.Finally,the neural network is used to identify characters.In order to verify the feasibility of the research method,the experimental study of each part is carried out and the experimental results are given.Then the whole system function is verified through data collected by millet UAV.The experimental results show that the research method can complete the function of license plate recognition in UAV video,and provide a reliable and effective solution for license plate recognition based on UAV video.