基于YOLOv5s和注意力机制的车辆检测改进算法

Improved Vehicle Detection Algorithm Based on YOLOv5s and Attention Mechanism

文中针对道路光线条件复杂、被遮挡目标物体特征不完整等问题,以YOLOv5s为车辆检测的基础模型,提出了一种融合FunelCBAM注意力机制的检测模型FCBAMYOLOv5s。针对交通场景中车辆检测种类多、多尺度目标混杂、小目标易漏检等问题,文中还提出了一种融合加权双向跨尺度特征金字塔BiFPN与KLLoss损失函数的车辆检测方法。该方法可融合多尺度图像问题的特征,提高检测图像的鲁棒性,强化了网络对小物体的检测性能;然后在检测损失函数中融合KLLoss,有效提高了模型的检测精度。实验结果表明,该方法的精度与实时性符合实际应用。

This paper problems of complex road light conditions and incomplete features of occluded target objects,this paper takes YOLOv5s as the basic model for vehicle detection,and proposes a detection model FCBAMYOLOv5s that in-tegrates the attention mechanism of Funel CBAM.Aiming at the problems of many types of vehicle detection in traffie scenes,mixed multi-scale targets,and easy missed detection of small targets,this paper also proposes a vehicle detection method that integrates weighted bidirectional cross-scale feature pyramid BiFPN and KL Loss loss function.This method can fuse the features of multi-scale image problems,improve the robustness of the detected image,and strengthen the de-tection performance of the network for small objects;then fuse KL Loss in the detection loss function,which effectively improves the detection accuracy of the model.Experimental results show that the accuracy and real-time performance of this method are in line with practical applications.