一种基于YOLOv5s的交通标志检测算法

Road Traffic Sign Detection Based on YOLOv5s

针对传统道路交通标志检测精度差、检测模型参数量较多等问题,对传统YOLOv5s模型进行了改进。该算法以YOLOv5s网络模型为基础,在骨干网络设计带有通道和空间注意力机制的C3CBAM卷积模块,从通道和空间域上增加对包含交通标志信息的特征图的关注度;接着在模型颈部将轻量级神经网络GhostNet融入原有的卷积网络,减少了网络模型参数量;最后在模型骨干网络和颈部增加跨层连接结构来融合更多的语义特征。将改进后的模型使用TT100K数据集进行训练和测试,实验结果表明改进后的模型参数量比之前减少了18.4%,且检测精度提升到了80.1%,比原有的检测网络提高了2.2%。

Aiming at the problems of poor accuracy and large number of detection model parameters in traditional road traffic sign detection,the traditional YOLOv5s model is improved.Based on the YOLOv5s network model,the algorithm designs a C3CBAM convolution module with channel and spatial attention mechanism in the backbone network,which increases the attention of the feature map containing traffic sign information from the channel and spatial domain.The lightweight neural network GhostNet is integrated into the original convolutional network at the model neck,which reduces the number of network model parameters.Finally,a cross-layer connection structure is added to the model backbone network and neck to integrate more semantic features.The experimental results show that the number of parameters of the improved model is reduced by 18.4%compared with the previous one,and the detection accuracy is increased to 80.1%,which is 2.2%higher than the original detection network.