Vehicle License Plate And Driver Face Detection Based On Deformable Parts Model

With the development of society,vehicles have becole the main means of transportation for people,and more and more serious road traffic safety problems have caused by vehicles.Therefore,the requirements for intelligent transportation systems are getting higher and higher.The license plate detection and driver face detection are important components of the intelligent traffic system and have become a hot issue in recent years.Due to the particularity of traffic road surveillance caleras,the license plate detection and driver face detection results are not satisfactory.In order to solve these problems,this paper proposes a vehicle license plate and driver face detection method based on deformable component model.The detection lethod designed in this thesis is divided into three phases:First,using the position information of the license plate,the license plate is detected from the image containing the vehicle and the license plate position information is given,and the position information of the license plate is used as a root template of the deformable component;Then,using the aggregate channel characteristics(LUV,local binary mode and direction gradient histogram),combined with AdaBoost classifier to detect some candidate face windows from the image to be detected.Finally,using the relative relationship between the license plate position and the drivers face position,the driver's face is used as a component template to filter false alarms by verifying the overall rationality of the license plate and the driver's face.During the training phase,we try to increase the detector recall rate and reduce the missed detection rate Then the false alarm is filtered out by the deformable component model,so that the accuracy rate increases,and the impact on the recall rate is small.The images we collected for traffic monitoring were divided into two sets of daytime and nighttime data sets,with a total of 2062 images.Using the data set to verify the validity of the proposed method,the driver's face detection accuracy during the day was 94.27%,and the recall was 86.17%.At night,the driver's face detection accuracy rate was 98.70%,and the recall rate was 96.21%.From the data,we can see that by adding the relative position features of the license plate and the human face,the accuracy of the driver's face detection is increased,and the detection effect of the driver face detection method used by us is ideal.