Robust Detection and Recognition of Japanese Traffic Sign in the Complex Scenes Based on Deep Learning
Ryô Hasegawa, Yutaro Iwamoto, Yen‐Wei Chen · 2019
This study applies the object detection and recognition method using deep learning, which has been attracting considerable attention recently, for the detection and recognition of traffic signs. Furthermore, the advantages and disadvantages of the proposed method are discussed. During the detection and recognition of traffic signs, the following problems were identified. (1) The detection and recognition accuracy of a small object in an image, such as a distant road sign, was low. (2) The accuracy of the traffic sign detection and recognition was affected due to the changes in contrast at night and bad weather. Herein, we address these problems by learning the traffic signs through deep learning, which is robust against scale changes. Herein, we construct a Japanese traffic sign database and compared the methods used. Results demonstrate that the proposed approach demonstrates excellent detection and recognition accuracy.