Research On Vehicle Face Feature Extraction And Recognition Method Based On Video Image

The core of Intelligent Transportation System(ITS)is the intelligent control on traffic objects,if you want to build a complete ITS,detection and identification technologies for vehicle in traffic scene is essential.So far,the public security system in the target vehicle monitoring,inspection work is mainly use the vehicle recognition technology based on the license plate to vehicle,and then get the information associated with the result which is based on license plate recognition,but in reality,there is a phenomenon that license plate is blocked,deface,not hanging or deck and so on.Therefore,for the vehicle license can’t be recognized,it is a difficult problem that how to use its shape features to recognize vehicle for narrowing the scope of the public security’s investigation work.Under this background,this paper research vehicle recognition technology which is based on vehicle’s face region feature,it has a very important significance in the public security investigation and traffic management.The research mainly includes the following four aspects:1.The improvement of the standard feature model library.For researching vehicle recognition based on the face region feature,establish a database table of standard feature is the key,the vehicle information including three categories:the physical characteristics,image characteristics and feature information in face image three,and take the physical characteristics of the class by the automatic web crawler way to improve real-time data,image feature class and vehicle’s face image feature class information is manually entered document library,it provides a reliable basis for classification and identification of the vehicle.2.Vehicle face location and region segmentation.According to Adaboost algorithm theory,put forward a algorithm with restraint experienced rectangle to improve traditional Adaboost algorithm and improved the vehicle face detection rate,then combined with the horizontal gradient and vertical gradient for vertical projection analysis,and effective located the radiator stall region,to analyze the textures characteristics of DCT and to achieve the purpose of positioning the car logo.3.The vehicle’s face region features extraction and analysis research.To extract face region image feature and study deeply including SIFT,SURF invariant features and LBP texture feature,and proposed an improved matching method based on SIFT,SURF feature recognition,then to select the right vehicle face region feature by feature matching recognition.Then analyzed Gabor feature on radiator stall,these provided the basis for the subsequent vehicle features face recognition.4.Research of vehicle face recognition based on multi-feature fusion.Combined with SIFT and SURF features of the vehicle to match the entire region of face for recognition,through the integration of Gabor frequency characteristic of radiator grille,proposed a algorithm for vehicle face recognition on multi-feature fusion,then for the vehicle models recognition of similar structures on vehicle face,proposed a model identification method that integrated face recognition classifier and vehicle logo recognition classifier.Statistics show that fusion of multi-car features face recognition can improve vehicle recognition rate.