A Novel Approach for Logo Recognition System Using Machine Learning Algorithm SVM

In different applications like Complex document image processing, Advertisement and Intelligent transportation logo recognition is an important issue. Logo Recognition is an essential sub process although there are many approaches to study logos in these fields. In this paper a robust method for recognition of a logo is proposed, which involves K-nearest neighbors distance classifier and Support Vector Machine classifier to evaluate the similarity between images under test and trained images. For test images eight set of logo image with a rotation angle of 0°, 45°, 90°, 135°, 180°, 225°, 270°, and 315° are considered. A Dual Tree Complex Wavelet Transform features were used for determining features. Final result is obtained by measuring the similarity obtained from the feature vectors of the trained image and image under test. Total of 31 classes of logo images of different organizations are considered for experimental results. An accuracy of 87.49% is obtained using KNN classifier and 92.33% from SVM classifier.