A sparse representation method for traffic sign recognition based on similar class
Yi Lei, Chongyang Zhang · 2014
In this paper, we propose a sparse representation method for traffic sign recognition based on similar class. The method needs to presort traffic signs as four main class according to its similar feature. We named the four main class as speed-limiting class, warning class, directive class and no-rules class. Then the method can be divided into two phases. First, we use a combination of PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) method to determine `the nearest neighbor' for the test sample. Second, we represent the test sample as a linear combination of the training samples from the main class that `the nearest neighbor' belongs to. Then we use the representation result to perform classification. Comparative experiments on German traffic signs database (GTSDB) show that the method is better than traditional methods such as OMP, PCA and LDA. Its recognition rate can reach 96%.