【摘要】 针对目前行人检测方法存在小目标检测难度大、漏检率高的问题，本文提出了一种改进YOLOv5s算法的行人检测方法。首先，增加一个小目标检测头，来增强模型对小目标的检测能力，根据自建数据集通过K-means聚类算法得到新的先验锚框尺寸；其次，将CA嵌入到YOLOv5s颈部网络的浅层位置和引入新型跨尺度特征融合模块加权特征融合来增强特征提取能力；最后，基于Ghost Bottleneck对YOLOv5s的C3模块进行改进，旨在通过低成本操作生成更多有价值冗余特征图，有效减少模型参数。实验结果表明，与原始YOLOv5s相比，改进的YOLOv5s算法在行人检测任务上的准确率P提高了3.3%，召回率R提高了2.9%,mAP＿0.5:0.95提高了2.6%，且减少了12.7%参数量，整体性能有显著提升。更多还原

【Abstract】 Aiming at the problems of difficult detection of small objects and high missed detection rate in the current pedestrian detection methods, this paper proposed a pedestrian detection method to improve YOLOv5s. Firstly, a small target detection head was added to enhance the detection ability of the model for small targets, and according to the self-built dataset, the new prior anchor frame size was obtained by the K-means clustering algorithm. Secondly, CA was embedded in the shallow position of YOLOv5s neck network and a new cross-scale feature fusion module weighted feature fusion was introduced to enhance the feature extraction ability. Finally, the C3 module of YOLOv5s was improved based on Ghost Bottleneck, aiming to generate more valuable redundant feature maps through low-cost operations and effectively reduce model parameters. The experimental results showed that, compared with the original YOLOv5s, the precision of the improved YOLOv5s in pedestrian detection tasks increased by 3.3 %, R increased by 2.9%, mAP＿0.5:0.95 increased by 2.6%, and the number of parameters was reduced by 12.7%, which was significantly improved.更多还原