基于改进 Faster R-CNN 的目标检测算法研究

doi:10.16638/j.cnki.1671-7988.2026.006.002

汽车实用技术 ›› 2026, Vol. 51 ›› Issue (6): 8-13.DOI: 10.16638/j.cnki.1671-7988.2026.006.002

• 自驾感知决策与控制 • 上一篇

基于改进 Faster R-CNN 的目标检测算法研究

曲雅婷，贾得顺

洛阳职业技术学院汽车与轨道交通学院

发布日期:2026-03-24
通讯作者: 曲雅婷
作者简介:曲雅婷（1993－），女，硕士，讲师，研究方向为智能网联汽车、车联网、信息与通信等
基金资助:
河南省科技厅重点研发推广专项（科技攻关）：基于数字孪生的自动驾驶场景生成与测试评价研究（252102240129）；校级科研项目：面向自动驾驶的目标检测关键技术研究（校 2024055）

Research on Object Detection Algorithm Based on Improved Faster R-CNN

QU Yating, JIA Deshun

School of Automobile and Rail Transit, Luoyang Polytechnic

Published:2026-03-24
Contact: QU Yating

摘要/Abstract

摘要： 针对自动驾驶车辆在复杂场景中的目标识别不精准的问题，文章提出了一种基于 Faster R-CNN 的改进型目标检测算法，采用残差网络（ResNet-50）来增强多尺度特征提取能力，优化锚框尺寸，借助多尺度卷积特征融合的方式来整合不同层次的特征，显著提升了目标检测模型对复杂场景的适应性。实验结果表明，改进型 Faster R-CNN 在整个召回率范围内均表现出更优的性能，尤其在高召回率区域依然能够保持较高的平均度均值（mAP）值，显示出良好的鲁棒性和泛化能力，验证了改进型 Faster R-CNN 算法的有效性，能够提升自动驾驶车辆在复杂场景中识别目标的精准度。

关键词: 自动驾驶；目标识别；Faster R-CNN；特征融合

Abstract: Aiming at the problem of inaccurate target recognition of autonomous vehicles in complex scenes, this paper proposes an improved target detection algorithm based on Faster R-CNN. In this paper, the residual network (ResNet-50) is used to enhance the ability of multiscale feature extraction, optimize the anchor frame size, and integrate different levels of features with the help of multi-scale convolution feature fusion, which significantly improves the adaptability of the target detection model to complex scenes. The experimental results show that the improved Faster R-CNN has better performance in the whole recall range, especially in the high recall area, it can still maintain a high mean average precision (mAP) value, showing good robustness and generalization ability, which verifies the effectiveness of the improved Faster R-CNN algorithm, and can improve the accuracy of automatic driving vehicles' target recognition in complex scenes.

Key words: autonomous driving; target recognition; Faster R-CNN; feature fusion

曲雅婷，贾得顺. 基于改进 Faster R-CNN 的目标检测算法研究[J]. 汽车实用技术, 2026, 51(6): 8-13.

QU Yating, JIA Deshun. Research on Object Detection Algorithm Based on Improved Faster R-CNN[J]. Automobile Applied Technology, 2026, 51(6): 8-13.

参考文献

[ 1 ] 曹家乐,李亚利,孙汉卿,等.基于深度学习的视觉目标检测技术综述[J].中国图象图形学报,2022,27(6): 1697-1722. [ 2 ] FU P,WANG J.Lithology Identification Based on Improved Faster R-CNN[J].Minerals,2024,14(9):954. [ 3 ] 程岩.基于智能网联汽车的目标识别融合算法研究 [J].时代汽车,2025(3):28-31. [ 4 ] 曹嘉禹,乔贵方,陈梦源,等.面向高空视角小目标检测的 YOLO 算法改进策略[J].计算机应用,2024,44 (S2):280-285. [ 5 ] HUANG X,DING K,LI B,et al.Remote Intelligent Identification of Suspended Foreign Objects in Substation Inspection Images Based on Edge Optimization Faster R-CNN Algorithm[J].Electrical Engineering, 2024,107(6):1-14. [ 6 ] ZANGANA M H,MUSTAFA M F,OMAR M. A Hybrid Approach for Robust Object Detection:Integrating Template Matching and Faster R-CNN[J].EAI Endorsed Transactions on AI and Robotics,2024,3(1):1-10. [ 7 ] 黄年昌,杨阳,张强,等.基于深度学习的 RGB-D 图像显著性目标检测前沿进展[J].计算机学报,2025,48 (2):284-316. [ 8 ] 刘桂超,王怀光,任国全,等.基于深度学习的单目视觉目标检测综述[J].计算机工程与应用,2026,62(1): 1-19. [ 9 ] OUF N S.Leguminous Seeds Detection Based on Convolutional Neural Networks:Comparison of Faster R-CNN and YOLOv4 on a Small Custom Dataset[J]. Artificial Intelligence in Agriculture,2023,8:30-45. [10] 罗元,王薄宇,陈旭.基于深度学习的目标检测技术的研究综述[J].半导体光电,2020,41(1):1-10. [11] 黄星威,陈曦,张塑凡.改进特征金字塔的小目标深度学习模型[J].计算机工程与科学,2023,45(4):734-742. [12] 邹逸群,肖志红,唐夏菲,等.Anchor-free 的尺度自适应行人检测算法[J].控制与决策,2021,36(2):295-302. [13] 汪常建,丁勇,卢盼成.融合改进 FPN 与关联网络的 Faster R-CNN 目标检测[J].计算机工程,2022,48(2): 173-179. [14] 陈憶悯,李万益,翁汉锐,等.基于深度学习的两阶段目标检测算法综述[J].信息与电脑(理论版),2023,35 (14):112-114. [15] 张琴,郭为安.深度学习小目标检测算法综述[J].计算机应用研究,2025,42(10):2893-2904.

基于改进 Faster R-CNN 的目标检测算法研究

Research on Object Detection Algorithm Based on Improved Faster R-CNN

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics