面向交通场景的轻量级行人检测算法

Lightweight Pedestrian Detection Algorithm for Traffic Scenes

针对交通场景下行人检测模型网络复杂、参数量大以及难以在低性能设备上部署的问题,基于YOLOv5s网络模型提出了一种改进的轻量级行人检测算法。首先,使用Ghost模块重构YOLOv5s网络进行特征提取,降低模型的参数量和计算量,提高推理速度。其次,引入坐标注意力机制提高模型对目标特征的提取能力,提升其对小目标行人的检测效果。最后,采用SIoU损失函数加快模型的收敛速度,提高模型的识别准确率。实验结果表明,改进后的算法能保证较高的检测精度,与原始YOLOv5s算法相比参数量减少47.1%,计算量减少48.7%,提高了交通场景下行人检测的速度且易于部署。

Aiming at the problems of pedestrian detection model in traffic scenes,such as complex network,large number of parameters,and difficulty in deploying on low-performance devices,an improved lightweight pedestrian detection algorithm based on YOLOv5s network model was proposed.Firstly,Ghost module was used to reconstruct the YOLOv5s network for feature extraction to reduce the number of parameters and computation of the model and improve the inference speed.Secondly,coordinate attention was introduced to improve the extraction ability of the model for target features and enhance the detection effect of small target pedestrians.Finally,the SIoU loss function was used to accelerate the convergence speed of the model and improve the recognition accuracy.The experimental results showed that the improved algorithm could reduce the number of parameters by 47.1%,and computational complexity by 48.7%compared to the original YOLOv5s algorithm,while ensuring high detection accuracy.The proposed algorithm could improve the speed of pedestrian detection in traffic scenes and was easy to deploy.