| The vehicle logo,which is a significant symbol of the vehicle and highly representative,cannot be easily damaged or tampered with.Vehicle logo recognition plays an essential role in the fields of intelligent transportation systems,smart cities,and intelligent driving.In the actual intelligent transportation system,due to fewer vehicles of some logos,the samples of those are difficult to obtain,and the collection of a large number of vehicle logo samples takes time and effort.Therefore,the vehicle logo recognition in the case of small samples is still of great significance.Given this,this paper studies the image of the vehicle logo obtained in the actual intelligent transportation system.Due to the small resolution of actual vehicle logo images and interference of the environment and various noises,the image quality is low.Therefore,this article attempts to study the vehicle logo recognition from the perspective of enhancing the features of the vehicle logo.This paper proposes two methods of vehicle logo feature enhancement,first,the vehicle logo recognition method based on HOL feature enhancement,second,the vehicle logo recognition method based on structural similarity and local feature enhancement.The main work of this article is as follows:(1)Analysis of existing vehicle logo recognition methods: The vehicle logo recognition methods based on traditional feature descriptors and the vehicle logo recognition methods based on deep learning are reviewed,meanwhile,related feature algorithms are briefly introduced and analyzed.(2)A vehicle logo recognition method based on HOL feature enhancement is proposed: combining the local feature information of the image,a method based on HOL feature enhancement is proposed,which can fully extract the global and local features of the vehicle logo image and enhance the image characteristic expression.In actual scenes,a higher recognition rate can be achieved with fewer training samples.(3)A vehicle logo recognition method based on structural similarity and local feature enhancement is proposed: combining the distinctive features of the image and the structural characteristics of the vehicle logo,a vehicle logo recognition method based on structural similarity and local feature enhancement is proposed.This method outperformed a mass number of experiments by extracting the overall feature of images,enhancing the expression of valid information from images with the local salient block feature and the self-symmetric similarity feature. |
PDF Full Text Request