A Novel SVM Network Using HOG Feature for Prohibition Traffic Sign Recognition

To recognize prohibition traffic sign, this paper proposes a novel method that is trained by small number of samples and uses the feature histogram oriented gradient (HOG) support vector machine (SVM) network. The recognition mainly divided into three stages. first stage image preprocessing, which includes interception based on ellipse detection, resizing, Gamma correction. In part interception, new detection called RHT_MCN proposed RHT, maximum coincidence (MCN) edge points detected to choose final for interception. second extraction HOG. third sign (PTSR) SVM design implementation PTSR model, single-layer network proposed. ascending spiral training model introduced in detail. Finally, data from GTSRB used test analyze method. proven have good applicability.