Research On Vehicle Recognition Technology Based On Binocular Vision

As an important part of intelligent transportation system,vehicle type identification plays an important role in the development of modern city.However,most of the existing ones are classified according to the vehicle size,vehicle standard,license plate,etc.,which can provide us with insufficient information.In many ways,we need to understand more detailed information about vehicle brands and models.At the same time,at present most of vehicle image acquisition from single camera,in some adverse environmental conditions,somespecial Angle of view,can't get a clear image of complete vehicle features extraction.Vehicle identification of a specific brand,model belongs to the class of problems in the field of fine-grained classification,and only according to different objects of image recognition,by contrast,for classification of vehicle brand,model similarity between high,not easy to distinguish.With the successful development of the convolution neural network in the field of image classification,the more effective recognition rate is compared with the traditional methods of feature extraction and pattern recognition in the past.In this paper,a complete model of vehicle brand and model is designed by combining the convolution neural networkand open source model data set with the label.This paper introduces the processing method of binocular vision at the same time,from the perspective of different position to vehicle image at the same time,studied including feature point extraction,feature point matching,the perspective transformation matrix parameters calculation,image fusion processing,etc.,so as to improve the accuracy of vehicle recognition in a complex environment.Through experimental verification,this method can effectively solve the problem of identifying the specific vehicle brand and model in a complex environment with limited single-vision Angle.