基于图像增强和改进RT-DETR的水下垃圾检测算法

doi:10.13474/j.cnki.11-2246.2025.0821

测绘通报 ›› 2025, Vol. 0 ›› Issue (8): 128-136.doi: 10.13474/j.cnki.11-2246.2025.0821

基于图像增强和改进RT-DETR的水下垃圾检测算法

李超^1,2, 刘清屹¹, 张佳伟³, 石勇¹, 杨敏³

1. 重庆市测绘科学技术研究院, 重庆 401120;
2. 自然资源部智能城市时空信息与装备工程技术创新中心, 重庆 401120;
3. 重庆交通大学机电与车辆工程学院, 重庆 400074

收稿日期:2025-01-03 出版日期:2025-08-25 发布日期:2025-09-02
作者简介:李超(1985—),男,硕士,正高级工程师,主要研究方向为自动化监测。E-mail:lichao@cqkcy.com
基金资助:
重庆市技术创新与应用发展专项面上项目(2023TIAD-GPX0097)

Underwater trash detection algorithm based on image enhancement and improved RT-DETR

LI Chao^1,2, LIU Qingyi¹, ZHANG Jiawei³, SHI Yong¹, YANG Min³

1. Chongqing Institute of Surveying and Mapping Science and Technology, Chongqing 401120, China;
2. Intelligent City Spatio-temporal Information and Equipment Engineering Technology Innovation Center of the Ministry of Natural Resources, Chongqing 401120, China;
3. Chongqing Jiaotong University Electromechanical Vehicles and Engineering College, Chongqing 400074, China

Received:2025-01-03 Online:2025-08-25 Published:2025-09-02

摘要/Abstract

摘要： 随着海洋环境污染问题的日益突出,水下垃圾的快速检测与清理尤为迫切。针对水下垃圾图像质量差、受光照影响严重、重叠和形状各异等导致检测效果差的问题,本文提出了一种基于改进RT-DETR的水下垃圾检测算法。针对图像存在色偏、对比度低等问题,设计了一种融合对比度增强和自适应色彩补偿的增强算法对图像进行预处理;针对移动设备模型轻量化的需求,引入FasterNet Block模块改进主干网络,减少模型参数量;针对水下环境光线弱的问题,采用HS-FPN高级筛选特征融合金字塔融合策略,解决特征损失严重和区分度低的问题;针对图像小目标居多,采用一种GELAN广义高效层聚合网络,提高模型的表征能力;针对空间位置造成垃圾尺寸差异大的问题,引入Inner-IoU与ShapeIoU结合的Inner-ShapeIoU损失函数,提高目标检测的稳健性。试验结果表明,本文方法有效解决了图像色偏和对比度低的问题,相较原模型检测精度提高3.9个百分点,参数量减少26.3个百分点,水下垃圾检测性能更加优越。

关键词: 水下垃圾检测, 深度学习, 图像增强, RT-DETR

Abstract: With the increasingly prominent problem of marine environmental pollution,rapid detection and cleaning of underwater garbage are particularly urgent.A new underwater garbage detection algorithm based on improved RT-DETR is proposed to address the issues of poor image quality,severe exposure to light,overlapping and varying shapes that lead to poor detection performance.Aiming at the problems of color cast and low contrast in images,an enhancement algorithm combining contrast enhancement and adaptive color compensation is designed for image preprocessing.In response to the demand for lightweight mobile device models,the FasterNet Block module is introduced to improve the backbone network and reduce the number of model parameters.To address the issue of weak lighting in underwater environments,the HS-FPN advanced filtering feature fusion pyramid fusion strategy is adopted to solve the problems of severe feature loss and low discrimination.For small targets in images,a GELAN generalized efficient layer aggregation network is adopted to improve the representation ability of the model.To address the issue of large differences in garbage size caused by spatial location,an Inner-ShapeIoU loss function combining Inner-IoU and ShapeIoU is introduced to improve the robustness of object detection.The experimental results show that the proposed method effectively solves the problems of image color cast and low contrast.Compared with the original model,the detection accuracy has been improved by 3.9 percent,and the number of parameters has been reduced by 26.3 percent.The underwater garbage detection performance is superior.

Key words: underwater trash detection, deep learning, image enhancement, RT-DETR

中图分类号:

P237

李超, 刘清屹, 张佳伟, 石勇, 杨敏. 基于图像增强和改进RT-DETR的水下垃圾检测算法[J]. 测绘通报, 2025, 0(8): 128-136.

LI Chao, LIU Qingyi, ZHANG Jiawei, SHI Yong, YANG Min. Underwater trash detection algorithm based on image enhancement and improved RT-DETR[J]. Bulletin of Surveying and Mapping, 2025, 0(8): 128-136.

参考文献

[1] 李陵云,王词稼,褚晓婷,等.我国海洋塑料垃圾治理体系与成效[J].海洋环境科学,2024,43(2):161-166.
[2] HIPOLITO J C,SARRAGA ALON A,AMORADO R V,et al.Detection of underwater marine plastic debris using an augmented low sample size dataset for machine vision system: a deep transfer learning approach[C]//Proceedings of the 19th IEEE Student Conference on Research and Development (SCOReD).Kota Kinabalu: IEEE,2021: 82-86.
[3] FULTON M,HONG J,ISLAM M J,et al.Robotic detection of marine litter using deep visual detection models[C]//Proceedings of 2019 International Conference on Robotics and Automation (ICRA).Montreal:IEEE,2019:5752-5758.
[4] TIAN Manjun,LI Xiali,KONG Shihan,et al.A modified YOLOv4 detection method for a vision-based underwater garbage cleaning robot[J].Frontiers of Information Technology & Electronic Engineering,2022,23(8):1217-1229.
[5] 袁红春,臧天祺.基于注意力机制及Ghost-YOLOv5的水下垃圾目标检测[J].环境工程,2023,41(7):214-221.
[6] ZHAO Yian,Lü Wenyu,XU Shangliang,et al.DETRs beat YOLOs on real-time object detection[EB/OL].[2024-12-23].https://arxiv.org/abs/2304.08069v3.
[7] PENG Y T,COSMAN P C.Underwater image restoration based on image blurriness and light absorption[J].IEEE Transactions on Image Processing,2017,26(4):1579-1594.
[8] WANG C Y,MARK LIAO H Y,WU Y H,et al.CSPNet:a new backbone that can enhance learning capability of CNN[C]//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).Seattle:IEEE,2020:1571-1580.
[9] WANG C Y,LIAO H M,YEH I H.Designing network design strategies through gradient path analysis[EB/OL].[2024-12-23].https://arxiv.org/abs/2211.04800v1.
[10] ZHANG H,ZHANG S.Shape-IoU:more accurate metric considering bounding box shape and scale[EB/OL].[2024-12-15].https://doi.org/10.48550/arXiv.2312.17663.
[11] HONG J,FULTON M,SATTAR J.TrashCan:a semantically-segmented dataset towards visual detection of marine debris[EB/OB].[2024-12-18].https://doi.org/10.48550/arXiv.2007.08097.
[12] PANETTA K,GAO Chen,AGAIAN S.Human-visual-system-inspired underwater image quality measures[J].IEEE Journal of Oceanic Engineering,2016,41(3):541-551.
[13] YANG Miao,SOWMYA A.An underwater color image quality evaluation metric[J].IEEE Transactions on Image Processing,2015,24(12):6062-6071.
[14] WANG Shiqi,MA Kede,YEGANEH H,et al.A patch-structure representation method for quality assessment of contrast changed images[J].IEEE Signal Processing Letters,2015,22(12):2387-2390.
[15] DREWS P,DO NASCIMENTO E,MORAES F,et al.Transmission estimation in underwater single images[C]//Proceedings of 2013 IEEE International Conference on Computer Vision Workshops.Sydney:IEEE,2013:825-830.
[16] KANG Yaozu,JIANG Qiuping,LI Chongyi,et al.A perception-aware decomposition and fusion framework for underwater image enhancement[J].IEEE Transactions on Circuits and Systems for Video Technology,2023,33(3):988-1002.
[17] ZHUANG Peixian,WU Jiamin,PORIKLI F,et al.Underwater image enhancement with hyper-Laplacian reflectance priors[J].IEEE Transactions on Image Processing,2022,31:5442-5455.
[18] WANG Huibai,WANG Yuxuan.Improved glove defect detection algorithm based on YOLOv5 framework[C]//Proceedings of the 6th IEEE Advanced Information Technology,Electronic and Automation Control Conference (IAEAC).Beijing:IEEE,2022:1192-1197.
[19] WANG C Y,BOCHKOVSKIY A,LIAO H M.YOLOv7:trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]//Proceedings of 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Vancouver:IEEE,2023:7464-7475.
[20] WANG C Y,YEH I H,LIAO H M.YOLOv9:learning what you want to learn using programmable gradient information[EB/OL].[2024-12-26].https://arxiv.org/abs/2402.13616v2.

基于图像增强和改进RT-DETR的水下垃圾检测算法

Underwater trash detection algorithm based on image enhancement and improved RT-DETR

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	胡耀锋, 程相兵, 陈嘉琦. 城市立体感知网建设关键技术与监测应用——以广州市历史文化街区为例[J]. 测绘通报, 2025, 0(8): 174-178.
[2]	张超, 熊春宝, 连思达. 基于深度学习的海洋平台基础环境监测方法[J]. 测绘通报, 2025, 0(6): 12-17.
[3]	闫白羽, 翟国君, 边少锋. 基于白化着色的沉船和失事飞机风格迁移方法[J]. 测绘通报, 2025, 0(6): 73-77.
[4]	陈登峰, 何拓航, 杨小燕, 刘世鹏, 孟屯良. 路基病害雷达图像多视图融合识别方法[J]. 测绘通报, 2025, 0(5): 66-73.
[5]	赵辰瑞, 连增增, 田亚林, 贺刘辉, 陈浩, 王鹏辉, 王孟奇. BiLSTM-Chan算法在超宽带室内定位中的应用[J]. 测绘通报, 2025, 0(5): 100-105.
[6]	朱松, 陈安泰, 华远盛, 姜文宇, 元鹏鹏, 朱家松. 面向漂流胶囊机器人图像的管道缺陷检测[J]. 测绘通报, 2025, 0(4): 34-38.
[7]	马于博, 张爱国, 王浩宇, 刘帅琪, 靳镜宇, 沈占锋, 李均力. 基于U-ConvHDNet模型的戈壁砾幕层提取[J]. 测绘通报, 2025, 0(4): 68-74.
[8]	李建, 王健, 王雷, 李敏, 杨立克, 赵艺龙. 双重注意力机制的电力走廊点云语义分割[J]. 测绘通报, 2025, 0(4): 127-133.
[9]	谢巴图, 胡佳睿, 潘俊. 基于双路全局信息优化网络的遥感影像海陆分割算法[J]. 测绘通报, 2025, 0(3): 52-58,86.
[10]	张璇, 杨本勇, 文武, 邓维熙, 周何帆. 改进的DeepLabV3+模型用于四川县域大小春作物识别[J]. 测绘通报, 2025, 0(3): 133-137,167.
[11]	苟长龙, 庞敏, 杨扬. 改进的U-Net卷积网络在遥感影像地物分类中的应用[J]. 测绘通报, 2025, 0(3): 150-155.
[12]	马海志. 基于深度学习的隧道超高分辨率图像病害检测框架[J]. 测绘通报, 2025, 0(3): 105-110.
[13]	宋培焱, 叶勤, 曾亮, 罗俊奇, 张硕, 尹长江. 面向密度明显差异点云的室内场景三维Mesh模型POCO重建方法[J]. 测绘通报, 2025, 0(2): 7-12,47.
[14]	石佳俊, 臧玉府, 肖雄武, 张莹滢. 非刚性匹配的激光点云与多光谱影像深度融合[J]. 测绘通报, 2025, 0(2): 41-47.
[15]	马浩然, 王金华. 多维特征学习与模型融合的地理命名实体识别[J]. 测绘通报, 2024, 0(9): 129-134.