The problem of recognizing signboards in street scenes is defined as matching the input image to pre-stored 2D signboard images. This problem is not as simple as it appears to be due to arbitrary drawings and relative 3D positions. We approach this problem by first matching local features of input image to those of images in the database. These matches are later verified with the global features, namely the nomographic consistency and the color consistency. The well-known SIFT feature is used as the local feature and the nomographic consistency checking is performed using RANSAC, a randomized method for finding an optimal match. In order to handle signboards with large perspective distortions, several templates with different perspective transformations are generated a-priori. In our experiments, the proposed method achieves up to 95 recognition rate, showing good results despite the highly distorted input images.
展开▼
