Image-based Vehicle Logo Location And Recognition

Vehicle logo recognition is a new research direction of vehicle identificationsystem, and it is also a hot topic of research in image processing, pattern recognitionand computer vision. So research of related technologies in this field has drawn wideattention. In this thesis, the related technologies of vehicle logo location andrecognition are studied, and an effective algorithm of vehicle logo location andrecognition is proposed. The main work includes following aspects:1. Existing vehicle logo location and recognition methods are studied, and theadvantages, disadvantages and limitations of these methods are summarized. Then wefind out the key points and difficulties of vehicle logo location and recognition, andpropose appropriate solutions.2. A vehicle logo location method based on prior knowledge and symmetrydetection is proposed. The prior knowledge refers to the position relationship amonglicense plate, vehicle logo and headlight area. Firstly, the edge character and colorfeature of license plate are utilized to accurately detect the license plate. Then, theheadlight area is located according to its edge feature and the position relationshipwith license plate. Finally, an advanced symmetry detection method is applied todetect the symmetry axis of the headlight area, and then the vehicle logo area islocated based on the position of the symmetry axis. The proposed method can locatethe vehicle logo in a slanted view image, and it will not be influenced by the complexbackground of the vehicle logo, so it has a wide applicable scope.3. A vehicle logo recognition method based on affine scale invariant featuretransform (ASIFT) is proposed. The query image is matched against the vehicle logotemplates respectively, and the template getting the maximum number of matches isconsidered to have the same type of vehicle logo with the query image. The ASIFTalgorithm is invariant to scale, translation and rotation, and it can deal with a widerange of viewpoints, so the vehicle logo recognition method based on the ASIFTalgorithm performs well for both frontal view images and slanted view images. It isproved that the proposed method achieves a high recognition rate.4. Software for vehicle logo location and recognition is realized using MicrosoftVC++6.0.