基于改进YOLOv7-Tiny的车辆检测研究  

Research on vehicle detection based on improved YOLOv7-Tiny

作　　者：李昊璇[1] 辛拓宇 LI Haoxuan;XIN Tuoyu(College of Physics and Electronic Engineering,Shanxi University,Taiyuan 030006,China)

机构地区：[1]山西大学物理电子工程学院,山西太原030006

出　　处：《电子设计工程》2025年第1期181-185,共5页Electronic Design Engineering

摘　　要：为了提高计算机识别检测车辆的准确度与速度,提出了一种基于改进YOLOv7-Tiny的车辆检测算法。在众多目标检测模型中,YOLOv7有着非常快的检测速度和较高的检测精度,非常适用于实时检测任务。在原YOLOv7-Tiny模型的基础上进行改进,将最浅层ELAN-T模块纳入特征金字塔,通过跳跃连接的方式将浅层特征与深层特征跨层融合,使输出的特征信息更加丰富。同时引入SE注意力机制,将计算资源分配给对当前任务更为关键的信息。并且更换了非线性激活函数HardSwish,以提高模型的表达能力。在华为发布的2D自动驾驶数据集SODA10M上进行实验,结果表明,改进后的模型对所有四种目标的检测精度都有所提高,平均精度mAP@0.5达到了66.1%,比原YOLOv7-Tiny模型61.0%提升了5.1%。In order to improve the accuracy and speed of computer recognition and vehicle detection,an improved YOLOv7⁃Tiny vehicle detection algorithm is proposed.Among numerous object detection models,YOLOv7 has a very fast detection speed and high detection accuracy,making it very suitable for real⁃time detection tasks.This article improves on the original YOLOv7⁃Tiny model by incorporating the shallowest ELAN⁃T module into the feature pyramid,and cross layer fusion of shallow and deep features is achieved through skip connections,resulting in richer output feature information.At the same time,the SE attention mechanism is introduced to allocate computing resources to information that is more critical to the current task.And the nonlinear activation function HardSwish was replaced to improve the model’s expression ability.Experiments were conducted on the 2D autonomous driving dataset SODA10M released by Huawei,and the results showed that the improved model showed improved detection accuracy for all four types of targets,with an average accuracy of mAP@0.5 reached 66.1%,an increase of 5.1%compared to the original YOLOv7⁃Tiny model of 61.0%.

关 键 词：目标检测 车辆检测 YOLOv7 注意力机制 深度学习 

分 类 号：TN911.73[电子电信—通信与信息系统]