Design and Implementation of intelligent road sign recognition system defending adversarial attack

• 1、School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876
• 2、School of economics and management,Beijing University of Posts and Telecommunications,Beijing 100876

Abstract：With the development of the Convolutional Neural Network, the recognition accuracy of image classification is becoming higher and higher. However, when facing adversarial images, the network has been demonstrated to be vulnerable. Our paper has designed and implemented an intelligent secure road sign recognition system, so as to defending adversarial attack. This work proposes a method which is to add spatial transformer networks(STNs) before the classification model. The STNs utilize the attention mechanism to extract the area of interest of the classification model and transform it to another vector space. Also, the adversarial training has been applied. These approaches maintain the basic structure informaDesign and Implementation of intelligent secure road sign recognition system based on defending adversarial attacktion of the original images against adversarial perturbations. Meanwhile, a presentation system based on Web has been developed. The experiments prove that the proposed methods are robust and effective at defending adversarial attacks.

Keywords： anti-adversarial attack road sign recognition adversarial training spatial transformer networks