基于改进YOLOv3的复杂环境下交通标志检测

针对深度学习算法中目标检测网络模型在复杂环境下识别交通标志的难点，对YOLOv3模型迁移学习算法的基本特点展开研究，构建并划分了复杂环境下中国交通标志数据集，并通过引入特征尺度的概念进一步改进YOLOv3算法，使数据集能够更好地处理各种复杂环境带来的影响。通过对比实验，证明改进后的YOLOv3算法对复杂环境下交通标志检测的效果明显优于标准YOLOv3算法及SSD算法，获得了更高的检测精度和更短的检测时间。

Traffic Sign Detection in Complex Environment Based on Improved Yolov3

Aiming at the difficulty of traffic signs recognition in complex environment by the target detection network model in the deep learning algorithm, the basic characteristics of the transfer learning algorithm of YOLOv3 model is studied, and the Chinese traffic sign data set in complex environment is constructed and divided. By introducing the concept of feature scale, YOLOv3 algorithm is further improved, so that the data set can better deal with the influence of various complex environments.Through comparative experiments, it is proved that the improved YOLOv3 algorithm is obviously superior to the standard YOLOv3 algorithm and SSD algorithm in traffic sign detection in complex environment,and achieves higher detection accuracy and shorter detection time.