基于改进YOLOv3的密集行人检测  

Dense pedestrian detection based on improved YOLOv3

作　　者：邓杰 万旺根[1,2] Deng Jie;Wan Wanggen(School of Communication and Information Engineering,Shanghai University,Shanghai 200072,China;Institute of Smart City,Shanghai University,Shanghai 200072,China)

机构地区：[1]上海大学通信与信息工程学院,上海200072 [2]上海大学智慧城市研究院,上海200072

出　　处：《电子测量技术》2021年第11期90-95,共6页Electronic Measurement Technology

基　　金：上海市科委港澳台科技合作项目(18510760300);中国博士后基金项目(2020M681264)资助。

摘　　要：行人检测是目标检测领域的一个重要分支,目前行人检测算法已经取得了较好的发展,但拥挤场景下存在着行人间的严重遮挡,这为检测任务带来了极大地挑战。为有效缓解该问题,在YOLOv3的基础上进行改进,提出单阶段密集行人检测算法:Crowd-YOLO,该算法将可见框标注信息加入到网络中,使网络同时预测全身框与可见框信息从而提升检测性能;提出时频域融合注意力模块(TFFAM),将频域通道注意力和空间注意力加入到网络中重新分配特征权重;采用数据关联型上采样代替传统的双线性插值,使深层特征图获取更为丰富的信息表达。使用非常具有挑战性的大型拥挤人群场景数据集CrowdHuman进行训练和测试,实验结果表明,所提方法比基础网络在AP50指标上提高了约3.7%,在召回率(Recall)指标上提高了3.4%,其中时频域融合注意力模块为网络带来了2.3%AP的性能增益。实验结果验证了所提方法在拥挤人群场景下的有效性。Pedestrian detection is an important branch in the field of object detection,and pedestrian detection algorithms have been well developed,but there exists severe occlusion between pedestrians in crowded scenes,which makes a great challenge for the detection task.To effectively alleviate this problem,this paper improves on YOLOv3 and proposes a single-stage dense pedestrian detection algorithm:Crowd-YOLO,which adds visible frame labeling information to the network to assist training,so that the network can predict both full-body frame and visible frame information to improve detection performance,proposes a time-frequency domain fused attention module(TFFAM),which adds frequency-domain channel attention and spatial attention to the network to redistribute features,uses data correlation upsampling instead of traditional bilinear interpolation to obtain a richer information representation of deep feature maps.This paper uses the very challenging large crowd scenario dataset CrowdHuman for training and testing.The experimental results show that the proposed method improves the AP50metric by about 3.7%and the recall metric by 3.4%over the baseline,with the time-frequency domain fused attention module bringing a 2.3%AP performance gain.The experimental results verify the effectiveness of the proposed method in crowded scenarios.

关 键 词：行人检测 遮挡问题 YOLO 融合注意力 数据关联型上采样 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TP332[自动化与计算机技术—计算机科学与技术]