Vehicle Model Identification

Abstract

Automobile has become one of the most important modes of transportation. The increasing number of automobiles has facilitated human life but it has also lead to various issues of traffic congestions, parking problems, traffic accidents etc. The objective of this project “Vehicle Model Identification” is to solve some of these problems. Vehicle Identification can be done by recognition of its iconic license plate (LP) but the Automatic License Plate Recognition (ALPR) System is rendered useless in case the license plate is forged, missing or covered. Another important attribute of a vehicle is its logo or emblem which contains important information about the car and as it cannot be tampered with easily, it plays an elemental role in classification and identification of vehicles. Vehicle Model Identification requires segmentation of vehicle logo from the given image followed by its recognition by matching it against a database of logos. The prerequisite for logo detection is the prior information about the License Plate (LP) area. Vehicle logo recognition is done using the feature matching approach provided by a robust image detector and descriptor called Speeded-Up Robust Features (SURF). In this project the experimentation has been confined to Light Motor Vehicles (LMVs) and under certain constrained conditions. Experiments on a number of downloaded as well self-acquired images of car were performed. An average accuracy of 82.5 % was obtained for logo detection using the modified algorithm and 61 % for logo recognition using SURF. The open source software OpenCV configured with CodeBlocks IDE http://sourceforge.net/projects/opencvlibrary/) has been used for experimentation.