Title :
An Improved Vehicle Logo Recognition Method for Road Surveillance Images
Author :
Quan Sun ; Xiaobo Lu ; Lin Chen ; Haihui Hu
Author_Institution :
Sch. of Autom., Southeast Univ., Nanjing, China
Abstract :
This paper presents an improved vision-based algorithm for detecting and recognizing vehicle logos in images captured by road surveillance cameras. Vehicle logo recognition is quite a challenging task considering the low resolution of the logos, the wide range of variability in illumination and the interference of the air-intake grille. However, our system, assessed on a set of 1386 vehicle images that belong to 15 distinctive vehicle manufactures, proves to be reliable and efficient under such circumstance. Firstly, we detect the location of the license plate using Adaboost and Local Binary Pattern (LBP) in order to reduce the searching area of the logo by prior knowledge, and then an improved gradient-based location algorithm is used to further focalize the logo target. Finally the logo is classified by Histograms of Oriented Gradients (HOG) and Support Vector Machine (SVM). Experimental results on a large number of images show the efficiency of the proposed scheme.
Keywords :
automobiles; character recognition; image classification; intelligent transportation systems; learning (artificial intelligence); support vector machines; surveillance; traffic engineering computing; Adaboost; HOG; LBP; SVM; air-intake grille; gradient-based location algorithm; histograms of oriented gradients; intelligent transportation system; license plate location detection; local binary pattern; logo classification; road surveillance cameras; road surveillance images; support vector machine; vehicle logo detection; vehicle logo recognition method; vision-based algorithm; Accuracy; Feature extraction; Image recognition; Image resolution; Licenses; Support vector machines; Vehicles; HOG features; Vehicle Logo Localization; Vehicle Logo Recognition; support vector machine;
Conference_Titel :
Computational Intelligence and Design (ISCID), 2014 Seventh International Symposium on
Print_ISBN :
978-1-4799-7004-9
DOI :
10.1109/ISCID.2014.12
