基于YOLOv8改进的室内行人跌倒检测算法FDW-YOLO
FDW-YOLO:An improved indoor pedestrian fall detection algorithm based on YOLOv8
陈晨 1徐慧英 1朱信忠 1黄晓 2宋杰 1曹雨淇 1周思瑜 1盛轲1
作者信息
- 1. 浙江师范大学计算机科学与技术学院(人工智能学院),浙江 金华 321004
- 2. 浙江师范大学教育学院,浙江 金华 321004
- 折叠
摘要
针对室内场景中由于光照变化、人体形态被遮挡以及在特殊视角下人体姿态变化等因素导致行人跌倒检测精度低、实时性差的问题,提出了一种基于YOLOv8改进的轻量级跌倒检测算法FDW-YOLO.将骨干网络中的C2f模块替换成FasterNext模块,增强有效特征复用的同时降低计算复杂度.根据人体跌倒姿势变化大的特点,设计了3种在颈部层不同位置添加动态可变形卷积模块的网络结构,并在自制的跌倒行为 目标检测数据集上进行实验比较,最终根据网络性能选择YOLOv8-C方案.在改进后的网络中引入边界框回归损失函数 WIoU取代原有的CIoU.实验结果表明,FDW-YOLO跌倒检测算法较YOLOv8n在mAP@0.5指标上从96.5％提升到97.9％,在mAP@0.5∶0.95指标上从72.5％提升到74.3％,同时参数量和计算量只有4.1×106和7.3×109,符合在低算力工业场景中部署的要求.
Abstract
Aiming at the problem of low fall detection accuracy and poor real-time performance in in-door scenes due to the effects of light change,occlusion of the human body form,and changes in the hu-man body posture under special viewpoint,a lightweight improved fall detection algorithm based on YOLOv8,named FDW-YOLO,is proposed.The C2f module in the backbone network is replaced by the FasterNext module,which reduces the computational complexity while retaining the excellent fea-ture extraction capability.According to the characteristics of human falls with large changes in posture,three network structures with dynamically deformable convolutional modules added at different positions in the neck layer are designed,experiments are conducted on a self-made fall dataset for comparison,and ultimately,the YOLOv8-C scheme is selected based on network performance.A bounding box re-gression loss function WIoU is introduced into the improved network to replace the original CIoU.The experimental results show that compared with YOLOv8n,the FDW-YOLO fall detection algorithm in-creases mAP@0.5 from 96.5％to 97.9％and mAP@0.5:0.95 from 72.5％to 74.3％,while the num-ber of parameters and computation is only 4.1 × 106 and 7.3 × 109,which is in line with the requirements for deployment in low-computing power industrial scenarios.
关键词
目标检测/跌倒/FasterNext/DDConv/WIoUKey words
object detection/fall/FasterNext/DDConv/WIoU引用本文复制引用
基金项目
国家自然科学基金(62376252)
国家自然科学基金(61976196)
浙江省自然科学基金重点项目(LZ22F030003)
国家级大学生创新创业训练计划项目创新训练重点项目(202310345042)
出版年
2024
国家科技期刊平台
NSTL
万方数据