基于TSD-YOLO的交通标志检测

Traffic Sign Detection Based on TSD-YOLO

摘要:

针对现有交通标志检测算法对小目标检测效果较差的问题,提出了一种基于级联多尺度特征融合的交通标志检测算法。首先,设计了一种新的级联多尺度特征融合网络,利用多尺度序列特征融合模块和三重特征编码模块,使模型能更好地融合交通标志的全局特征和细节特征。其次,在骨干网络中加入可变形注意力机制,使模型专注于相关区域并捕获更丰富的图像特征。最后,使用Inner-IoU 损失函数,提升了模型的泛化性能。在CCTSDB数据集上的测试结果表明,改进模型的平均精度为55.3%,较YOLOv8s模型提升了3.2%。此外,在TT100K和VOC数据集上的表现凸显了模型出色的泛化性能。

Abstract:

A multi-scale feature fusion based traffic sign detection algorithm is proposed to address the problem of poor performance of existing object detection algorithms in detecting small target traffic signs. Firstly, a new cascaded multi-scale feature fusion network was designed, which fully utilizes the multi-scale sequence feature fusion structure and triple feature encoding module, enabling the network to better integrate the detailed and global features of traffic signs. Secondly, incorporating deformable attention mechanisms into the backbone network enables the model to focus on relevant regions and capture richer image features. Finally, the use of the Inner IoU loss function improved the generalization performance of the model. The experimental results on the CCTSDB dataset show that the average accuracy of the improved model is 55.3%, which is 3.2% higher than YOLOv8s. In addition, the performance on the TT100K and VOC datasets highlights the excellent generalization performance of the model.