激光与光电子学进展, 2019, 56 (19): 191002, 网络出版: 2019-10-12
基于改进神经网络的交通标志识别 下载: 1099次
Traffic Sign Recognition Based on Improved Neural Networks
图像处理 神经网络 交通标志识别 YOLOv2 损失函数 image proceedings neural network traffic sign recognition YOLOv2 loss function
摘要
交通标志识别在驾驶辅助系统和交通安全方面发挥着重要作用。卷积神经网络在计算机视觉任务上取得了重大的突破,并在交通标志检测与识别方面取得了巨大的成功。然而,现有的识别方法通常达不到实时识别的效果。因此,提出一种改进卷积神经网络交通标志识别方法,通过加入初始模块,扩展网络结构和提出新的损失函数等多种方法来解决原始模型不擅于检测小目标的问题。在德国交通标志数据集上的仿真结果表明,与现有技术相比,提出的方法能够获得更高的检测速率,每张图片的处理时间仅为0.015 s。
Abstract
Traffic sign recognition plays an important role in driver assistance systems for traffic safety. Convolutional neural networks (CNNs) have made a significant breakthrough in computer vision tasks and achieved considerable success in traffic sign detection and recognition. However, existing methods typically fail at achieving real-time recognition. Therefore, this study proposes a modified traffic sign recognition method based on a CNN, wherein inception modules are added, the network structure is extended, and a new loss function is used to overcome the original model's difficulty in detecting small targets. German traffic sign datasets are used to simulate the effectiveness of the proposed method. Simulation results show that the proposed method can obtain higher detection rates than those of existing methods at the processing time of only 0.015 s for each image.
童英, 杨会成. 基于改进神经网络的交通标志识别[J]. 激光与光电子学进展, 2019, 56(19): 191002. Ying Tong, Huicheng Yang. Traffic Sign Recognition Based on Improved Neural Networks[J]. Laser & Optoelectronics Progress, 2019, 56(19): 191002.