增强小目标检测的建筑工地安全装备检测.pdfVIP

计算机系统应用

ISSN

1003-3254,

CODEN

CSAOBNE-mail:

csa@

2025,34(2):122−134

[doi:

10.15888/ki.csa.009751]

[CSTR:

32024.14.csa.009751]

©中国科学院软件研究所版权所有.Tel:

+86-10

增强小目标检测的建筑工地安全装备检测①

吕明海,

王昱博,

吕伏,

冯永安

(辽宁工程技术大学

软件学院,

葫芦岛

125105)

通信作者:

王昱博,

E-mail:

984006976@

摘要:

YOLOv8n算法在面对背景繁杂、目标密集、像素点小的情况下,

表现出识别精度欠佳、目标漏检及误识

别的问题.

针对上述问题,

提出一种LNCE-YOLOv8n安全装备佩戴检测算法.

包括提出线性多尺度融合注意力

LMSFA

(linear

multi-scale

fusion

attention)机制,

自适应聚焦关键特征,

提升对小目标信息提取的能力且减少计算.

提出C2f_NewNet

(C2f_New

network)结构,

通过有效的并行化设计,

保持高性能且减少深度.

结合轻量级通用上采

样算子CARAFE

(content-aware

reassembly

of

feature),

实现跨尺度的高效特征融合与传播,

在大的感受野内聚合上

下文信息.

基于SIoU

(symmetric

intersection

over

union)损失函数提出ESIoU

(enhanced

symmetric

intersection

over

union),

提升模型在复杂环境中的适应性和精度.

实验采用safety

equipment数据集进行训练测试,

结果表明LNCE-

YOLOv8n算法相比YOLOv8n算法,

精度提升了5.1%,

mAP50提升了2.7%,

mAP50-95提升了3.4%,

有效提高建筑

工地复杂场景的工人安全装备佩戴检测精度.

关键词:

YOLOv8n;

注意力机制;

并行化设计;

上采样;

损失函数

引用格式:

吕明海,王昱博,吕伏,冯永安.增强小目标检测的建筑工地安全装备检测.计算机系统应用,2025,34(2):122–134.

/

1003-3254/9751.html

ConstructionSiteSafetyEquipmentDetectionwithEnhancedSmallTargetDetection

LYU

Ming-Hai,

WANG

Yu-Bo,

LYU

Fu,

FENG

Yong-An

(Software

College,

Liaoning

Technical

University,

Huludao

125105,

China)

Abstract:

The

YOLOv8n

algorithm

exhibits

suboptimal

performance

when

dealing

with

complex

backgrounds,

dense

targets,

and

small-sized

objects

with

limited

pixel

information,

leading

to

reduced

precision,

missed

detection,

and

misclassification.

To

address

these

issues,

this

study

proposes

an

algorithm,

LNCE-YOLOv8n,

for

safety

equipment

detection.

This

algorithm

includes

a

linear

multi-scale

fusion

attention

(LMSFA)

mechanism,

which

adaptively

focuses

on

key

features

to

improve

the

extraction

of

information

from

small

targets

wh