The method of recognizing traffic signs based on the improved capsule network

Abstract

Traffic sign recognition is one of the urgent problems to be solved by automatic driving technology, and it is also one of the more complex problems. For the problem that the conventional convolutional neural network has not been good enough to recognize traffic signs, this paper uses an improved capsule network .The method first uses image processing to extract features of traffic signs in a complex background, remove noise, binarize traffic signs, extract the main parts, make the characteristics of traffic signs more obvious, and then input the traffic signs into the capsule network to identify. The test results on the GTSRB data set show that the improved capsule network method has an improved recognition accuracy of 2%-5% in complex scenes, which is a great improvement compared to the traditional convolutional neural network. The experimental results show that the improved capsule network method has great reference significance for the research of autonomous driving.