The Vehicle Logo Recognition Software Based On SIFT Algorithm

With the rapid economic development of China and the improvement of people’smaterial and cultural standard, people’s demand for cars is increasing. Facing the problem ofmore and more cars, how to effectively manage these vehicles becomes a primary concern.Logo is an important feature of the vehicle. Compared to some other feature information(such as license plate), vehicle logo is not easy to be changed. Based on the important feature,this paper will do a research on the recognition of the vehicle logo.The paper is divided into two core parts, one is positioning part (including the licenseplate positioning and vehicle logo positioning). The other is the extraction of the featurevector. The logo positioning part is based on the prior knowledge between logo and vehiclelicense plate, using a coarse-to-fine vehicle logo positioning methods. First, positioning thelicense plate through the mature methods, and then estimating the logo area through relativeposition between logo and license plate.Then using SOBLE edge detection operator detectingthe vertical edge and horizontal edge on the vehicle logo at the same time.Separating thevehicle logo area and background area.Then use mathematical morphology to corrode andexpand the logo area, and obtain accurate vehicle logo area. Vehicle logo feature vectorextraction and recognition is based on sift algorithm in this paper. The main advantage ofSIFT operator is informative, unique and suitable for accurate matching in the massdata.Another advantage is that SIFT operator is local features.so it remains invariant inluminance variation, the scale of the image and the rotation. It also maintains a high stabilityin the change of viewing angle, radiation transformation and noise.In addition, SIFT operatoris scalable and can produce more feature points even within small number of objects.It is alsoconvenient to join with other forms feature vector. Therefore, this paper will adopt SIFToperator for the extraction and recognition of feature vector. And finally through theEuclidean distance, judging the feature vector similarity, ultimately recognizing the vehiclelogo.If the vehicle logo recognition technology is applied to criminal investigation, it willgreatly reduce the search scope of the crime and relevant departments work load and improvethe reliability of the vehicle recognition.