Traffic Sign Recognition based on multi-block LBP features using SVM with normalization

The automatic traffic sign detection and recognition has been converted to a real challenge for high performance of computer vision and machine learning techniques. It is an important issue, in particular for vehicle safety applications. It is usually tackled in three stages: detection, feature extraction and classification. We focus in this work on the second stage of the process, namely traffic sign feature extraction with applying the Block-Based Local Binary Pattern (LBP) method. This paper investigates the use of two normalization techniques: min-max and z-score. These techniques are incorporated in the SVM prediction model. For comparison purposes we also implemented the block-Based LBP with the K-Nearest Neighbor (K-NN) classifier. The evaluation of the proposed approach on the German Traffic Sign Recognition Benchmark Dataset (GTSRD) proved that the normalization of SVM input space can significantly influence the higher accuracy performance of the classification procedure. Extensive and comparative experiments have been conducted to evaluate our proposed method.