传感技术学报2026，Vol.39Issue(3)：582-590,9.DOI:10.3969/j.issn.1004-1699.2026.03.015
改进YOLOv7的自动驾驶目标检测算法
An Autonomous Driving Object Detection Algorithm Based on Improved YOLOv7
摘要
Abstract
To address the challenges of perception and detection accuracy in complex traffic environments,an enhanced YOLOv7 model is proposed to meet the demands of multi-target detection.Through feature augmentation techniques,the original network structure is op-timized to achieve multi-scale feature fusion,significantly boosting the model's feature representation capabilities.Additionally,the intro-duction of the GE attention mechanism further enhances the extraction of multi-scale features,effectively improving target detection ac-curacy.Furthermore,the integration of CoordConv in the neck and detection head of the model significantly enhances the network's abil-ity to capture spatial information,optimizing its learning capabilities and performance.The improved YOLOv7 achieves an average preci-sion of 48.2%,an increase of 5.6 percentage points,and a recall rate improvement of 9.2%.These results demonstrate that the enhanced algorithm is capable of meeting the requirements of target detection in autonomous driving's complex environments.关键词
目标检测/YOLOv7/注意力机制/多尺度特征网络/自动驾驶Key words
target detection/YOLOv7/attention mechanism/multi-scale feature network/autonomous driving引用本文复制引用
江自豪,杨思远,王世康,王坤相,何宇豪,王冠凌..改进YOLOv7的自动驾驶目标检测算法[J].传感技术学报,2026,39(3):582-590,9.基金项目
国家自然科学基金项目(U22A2079) （U22A2079）
安徽高校自然科学研究重大项目(J2021ZD0116) （J2021ZD0116）
皖江高端装备制造协同创新中心开放基金项目(GCKJ2018007) （GCKJ2018007）
