融合滑动窗口与改进SGBM的高分辨率目标检测与测距研究

丁建军; 汪刚林; 陈正超; 张超; 夏礼恒; 夏骏麒; 孙超

doi:null

您当前的位置：

首页 >

文章列表页 >

融合滑动窗口与改进SGBM的高分辨率目标检测与测距研究

光学传感与测量 | 更新时间：2026-01-22

- 融合滑动窗口与改进SGBM的高分辨率目标检测与测距研究
- High-resolution object detection and ranging via sliding window integration and enhanced SGBM
- 光学技术 2026年52卷第1期页码：86-93
- 作者机构：
  
  1.江汉大学专用装备智能测控湖北省工程研究中心，湖北武汉 430056
  2.江汉大学智能制造学院，湖北武汉 430056
- 作者简介：
  
  丁建军（1969— ），男，教授，博士，从事化学传感器、智能检测、智能制造、图像处理、机电控制研究。10107093@jhun.edu.cn
  孙超（1984— ），男，副教授，博士，从事传感器设计、深度学习算法、计算机视觉研究。sycsc115@163.com
- 基金信息：
  
  国家重点研发计划基金资助项目(2018YFD1100104);湖北省教育厅教学研究项目(2022277┫武汉市教育局产学研项目┣CXY202210)
- DOI：
  中图分类号： TP394.1;TH691.9
- 收稿：2025-05-30，
  
  修回：2025-08-24，
  
  纸质出版：2026-01-20
- 稿件说明：
移动端阅览
丁建军,汪刚林,陈正超等.融合滑动窗口与改进SGBM的高分辨率目标检测与测距研究[J].光学技术,2026,52(01):86-93.

DING Jianjun,WANG Ganglin,CHEN Zhengchao,et al.High-resolution object detection and ranging via sliding window integration and enhanced SGBM[J].OPTICAL TECHNIQUE,2026,52(01):86-93.
丁建军,汪刚林,陈正超等.融合滑动窗口与改进SGBM的高分辨率目标检测与测距研究[J].光学技术,2026,52(01):86-93. DOI：

DING Jianjun,WANG Ganglin,CHEN Zhengchao,et al.High-resolution object detection and ranging via sliding window integration and enhanced SGBM[J].OPTICAL TECHNIQUE,2026,52(01):86-93. DOI：

摘要

随着成像技术的发展，高分辨率图像广泛应用于多个领域，但在复杂天气、多尺度变化和立体匹配复杂度高等条件下仍面临挑战。提出一种结合滑动窗口的目标检测与测距方法，在处理高分辨率图像时，采用滑动窗口法对图像进行分割并保留20%的重叠区域，随后将各子图输入至YOLOv8模型进行检测。为提升模型性能，在backbone中引入GCANet模块以增强雾霾环境下的图像质量，并加入HorBlock以提升多尺度特征表征能力。检测完成后，合并各子图结果，并采用引入了CIoU匹配机制的非极大值抑制算法（NMS）进行后处理，以缓解密集目标场景中的过抑制问题。最后，设计了基于检测框约束的局部匹配算法，结合加权最小二乘滤波（WLS）补全视差空洞。实验结果表明，在自制的雾天场景Fog-A数据集与公开的无人机视角VisDrone2019数据集上，方法相较于原始YOLOv8n模型，mAP分别提升9.8%和6.1%；匹配时间缩短至原SGBM的三分之一，精度提升0.7%。

Abstract

With the rapid advancement of imaging technologies

high-resolution images have found widespread applications across diverse fields.Nevertheless

challenges persist under complex weather conditions

multi-scale variations

and the high computational burden of stereo matching. A high-resolution object detection and distance measurement framework that integrates a sliding window strategy is proposed. Specifically

input images are partitioned into sub-images with 20% overlap and subsequently processed by the YOLOv8 model for detection.To enhance model robustness

a GCANet（Gated Context Aggregation Network） module is incorporated into the backbone to enhance feature robustness against atmospheric degradation

while a HorBlock（High-order Interaction Block） module is introduced to strengthen multi-scale feature representation.After detection

results from the sub-images are merged

and a CIoU-based NMS is applied to alleviate over-suppression in dense object scenarios.Furthermore

a detection box-constrained local stereo matching algorithm is designed

which is then refined with Weighted Least Squares （WLS） filtering to fill disparity holes and improve depth estimation.Experimental results on a self-constructed foggy dataset （Fog-A） and the public VisDrone2019 dataset show that the proposed method achieves mAP improvements of 9.8% and 6.1%

respectively

over the baseline YOLOv8n model. Moreover

the proposed local matching algorithm reduces computation time to one-third of that of the original SGBM and improves matching accuracy by 0.7%.

关键词

Keywords

references

刘家晨 . 高分辨率图像中的小目标识别算法研究与实现 [D ] . 武汉 : 华中科技大学 , 2020 .

Liu Jiachen . Research and implementation of small object detection algorithms in high-resolution images [D ] . Wuhan : Huazhong University of Science and Technology , 2020 .

周璇 , 葛琦 , 邵文泽 . 高分辨率特征增强的无人机航拍小目标检测 [J ] . 数据采集与处理 , 2024 , 39 （ 04 ）: 908 — 921 .

Zhou Xuan , Ge Qi , Shao Wenze . Small object detection in UAV aerial images with high-resolution feature enhancement [J ] . Data Collection and Processing , 2024 , 39 （ 04 ）: 908 — 921 .

Hoanh N , Pham T V . Focus-attention approach in optimizing detr for object detection from high-resolution images [J ] . Knowledge-Based Systems , 2024 , 296 : 111939 .

尚柯言 . 基于深度学习的无人艇海面场景目标检测方法研究 [D ] . 长沙 : 湖南大学 , 2023 .

Shang Keyan . Research on target detection methods for unmanned surface vehicle maritime scenes based on deep learning [D ] . Changsha : Hunan University , 2023 .

Zhou Y , Jing W , Wang J , et al . MSAR‐ DefogNet:Lightweight cloud removal network for high resolution remote sensing images based on multi scale convolution [J ] . IET Image Processing , 2022 , 16 （ 3 ）: 659 — 668 .

Wang F , Wang X . HSF-YOLO:a multi-scale and gradient-aware network for small object detection in remote sensing images [J ] . Sensors （Basel,Switzerland） , 2025 , 25 （ 14 ）: 4369 .

原铭 , 李凡 , 李华锋 , 等 . 基于多级注意力密集残差网络的单幅图像超分辨率重建 [J ] . 光学技术 , 2022 , 48 （ 03 ）: 357 — 363 .

Yuan Ming , Li Fan , Li Huafeng , et al . Single-image super-resolution reconstruction based on multi-level attention dense residual network [J ] . Optical Technique , 2022 , 48 （ 03 ）: 357 - 363 .

李新宇 . 基于深度学习的高分遥感图像建筑物变化检测研究 [D ] . 大连 : 大连交通大学 , 2025 .

Li Xinyu . Research on building change detection in high-resolution remote sensing images based on deep learning [D ] . Dalian : Dalian Jiaotong University , 2025 .

Du D , Zhu P , Wen L , et al . VisDroneDET2019:The vision meets drone object detection in image challengeresults [C ] // Proceedings of The IEEE/CVF International Conference on Computer Vision Workshops . Long Beach,USA : IEEE , 2019 : 213 — 216 .

Bodla N , Singh B , Chellappa R , et al . Soft-NMS-improving object detection with one line of code [C ] // Proceedings of The IEEE International Conference on computer Vision . Venice,Italy : IEEE , 2017 : 5561 — 5569 .

Du S , Zhang B , Zhang P , et al . An improved bounding box regression loss function based on CIOU loss for multi-scale object detection [C ] // 2021 IEEE 2nd International Conference on Pattern Recognition and Machine Learning （PRML） . Chengdu,China : IEEE , 2021 : 92 — 98 .

Zheng Z , Wang P , Liu W , et al . Distance-IoU loss:Faster and better learning for bounding box regression [C ] // Proceedings of The AAAI Conference on Artificial Intelligence . New York,USA : AAAI , 2020 : 12993 — 13000 .

Zhai H , Cheng J , Wang M . Rethink the IoU-based loss functions for bounding box regression [C ] // 2020 IEEE 9th Joint International Information Technology and Art Ificial Intelligence Conference （ITAIC） . Chongqing,China : IEEE , 2020 : 1522 — 1528 .

Das S , Islam M S , Amin M R . Gca-net:utilizing gated context attention for improving image forgery localization and detection [C ] // Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans,USA : IEEE , 2022 : 81 — 90 .

Rao Y , Zhao W , Tang Y , et al . Hornet:Efficient high-order spatial interactions with recursive gated convolutions [J ] . Advances in Neural Information Processing Systems , 2022 , 35 : 10353 — 10366 .

Wang J , Hu X . Convolutional neural networks with gated recurrent connections [J ] . IEEE Transactions on Pattern Analysis and Machine Intelligence , 2021 , 44 （ 7 ）: 3421 — 3435 .

许鸿萍 . 基于机器视觉的工业机器人相机标定方法研究 [J ] . 现代制造技术与装备 , 2024 , 60 （ 12 ）: 209 — 211 .

Xu Hongping . Research on camera calibration method of industrial robots based on machinevision [J ] . Modern Manufacturing Technology and Equipment , 2024 , 60 （ 12 ）: 209 — 211 .

吕钦 , 周诗洋 , 万波为 , 等 . 基于双目视觉的铝合金件三维点云重建研究 [J ] . 组合机床与自动化加工技术 , 2025 ,（ 02 ）: 63 — 68 .

Lv Qin , Zhou Shiyang , Wan Bowei , et al . Research on 3D point cloud reconstruction of aluminum alloy parts based on binocular vision [J ] . Modular MachineTool & Automatic Manufacturing Technique , 2025 ,（ 02 ）: 63 — 68 .

Liu Y , Huang W , Wang Y . A new stereo matching method for RAW image data based on improved SGBM [C ] // Three-Dimensional Image Acquisition and Display Technology and Applications . Beijing,China : SPIE , 2018 , 10845 : 208 — 214 .

Cai Zhenhao , Yu Hongyi . Optimal design of hybridfilter bank based on weighted least squares regularization method [J ] . Journal of Signal Processing , 2010 , 26 （ 7 ）: 1110 — 1114 .

Zhu X , Su W , Lu L , et al . Deformable DETR: deformable transformers for end-to-end object detection [C ] // International Conference on Learning Representations Vienna , Austria : Open-Review.net . 2021 : 1 — 16 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

具有大成像深度高层析能力的高分辨率共聚焦显微物镜设计

多光纤阵列输入高分辨率微型拉曼光谱仪设计