针对浒苔目标检测的全局背景强化的位置蒸馏方法

doi:10.13474/j.cnki.11-2246.2024.0604

测绘通报 ›› 2024, Vol. 0 ›› Issue (6): 19-23.doi: 10.13474/j.cnki.11-2246.2024.0604

针对浒苔目标检测的全局背景强化的位置蒸馏方法

刘兵^1,2, 刘宇², 金凤学², 邹一波^1,3, 葛艳³, 赵林林³

1. 自然资源部生态预警与保护修复重点实验室, 山东青岛 266033;
2. 三峡新能源盐城大丰有限公司, 江苏盐城 224199;
3. 上海海洋大学信息学院, 上海 201308

收稿日期:2023-10-05 发布日期:2024-06-27
作者简介:刘兵(1979—),男,硕士,高级工程师,研究方向为图像处理。E-mail:liu_bing@ctg.com.cn
基金资助:
自然资源部生态预警与保护修复重点实验室开放基金(2022105)

Distillation method based on global background strengthen for Enteromorpha prolifera target detection

LIU Bing^1,2, LIU Yu², JIN Fengxue², ZOU Yibo^1,3, GE Yan³, ZHAO Linlin³

1. Key Laboratory of Ecological Prewarning, Protection and Restoration of Bohai Sea, Ministry of Natural Resources, Qingdao 266033, China;
2. Three Gorges New Energy Yancheng Dafeng Co., Ltd., Yancheng 224199, China;
3. The School of Information, Shanghai Ocean University, Shanghai 201308, China

Received:2023-10-05 Published:2024-06-27

摘要/Abstract

摘要： 浒苔检测是目前海洋环境智能监测领域研究的重要课题之一。为了有效解决传统浒苔检测方法存在的训练样本需求大的问题,本文提出了一种全局背景强化的位置蒸馏模型(GBS-LD)。通过引入全局上下文模块和背景蒸馏损失分支,解决了原始位置蒸馏方法在建模背景特征上的不足,在复杂海洋环境下有效提高了浒苔检测系统的稳健性。在浒苔检测数据集中,本文模型具有较高的准确性和实时性,为海洋智能监测提供了重要参考。

关键词: 浒苔, 位置蒸馏, 全局背景强化, 目标检测, 深度学习

Abstract: The detection of Enteromorpha prolifera stands as a pivotal research area within the realm of intelligent marine environment monitoring.Addressing the challenge posed by the substantial training sample requirements inherent in conventional methods of Enteromorpha prolifera detection,this paper proposes GBS-LD model.By introducing a global context module and background distillation loss branch,the shortcomings of the original position distillation method in modeling background features are solved,effectively improving the robustness of detection system in complex marine environments.Our proposed model has achieved high accuracy and real-time performance in the dataset of Enteromorpha prolifera,providing important reference for intelligent monitoring of marine.

Key words: Enteromorpha prolifera, localization distillation, global background strengthing, object detection, deep leaning

中图分类号:

P237

刘兵, 刘宇, 金凤学, 邹一波, 葛艳, 赵林林. 针对浒苔目标检测的全局背景强化的位置蒸馏方法[J]. 测绘通报, 2024, 0(6): 19-23.

LIU Bing, LIU Yu, JIN Fengxue, ZOU Yibo, GE Yan, ZHAO Linlin. Distillation method based on global background strengthen for Enteromorpha prolifera target detection[J]. Bulletin of Surveying and Mapping, 2024, 0(6): 19-23.

参考文献

[1] 王怡人,王胜强,喻樾,等.一种提取南黄海浒苔的自适应阈值遥感算法[J].遥感信息,2021,36(2):120-129.
[2] 麻德明,张秀林,田梓文,等.无人机遥感影像浒苔信息提取方法[J].测绘通报,2023(8):24-28.
[3] WANG Shengke,LIU Lu,QU Liang,et al.Accurate Ulva prolifera regions extraction of UAV images with superpixel and CNNs for ocean environment monitoring[J].Neurocomputing,2019,348(C):158-168.
[4] GAO Le,LI Xiaofeng,KONG Fanzhou,et al.AlgaeNet:a deep-learning framework to detect floating green algae from optical and SAR imagery[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2022,15:2782-2796.
[5] JIN Xifang,LI Yun,WAN Jianhua,et al.MODIS green-tide detection with a squeeze and excitation oriented generative adversarial network[J].IEEE Access,2022,10:60294-60305.
[6] HINTON G E,VINYALS O,DEAN J.Distilling the knowledge in a neural network[EB/OL].[2015-03-09].https://arxiv.org/abs/1503.02531.
[7] ZHENG Zhaohui,YE Rongguang,WANG Ping,et al.Localization distillation for dense object detection[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).New Orleans:IEEE,2022:9397-9406.
[8] LI X,WANG W,WU L,et al.Generalized focal loss:learning qualified and distributed bounding boxes for dense object detection[J].Advances in Neural Information Processing Systems,2020,33:21002-21012.
[9] TIAN Zhi,SHEN Chunhua,CHEN Hao,et al.FCOS:fully convolutional one-stage object detection[C]//Proceedings of 2019 IEEE/CVF International Conference on Computer Vision (ICCV).Seoul:IEEE,2019:9626-9635.
[10] GUO Jianyuan,HAN Kai,WANG Yunhe,et al.Distilling object detectors via decoupled features[C]//Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Nashville:IEEE,2021:2154-2164.
[11] YANG Zhendong,LI Zhe,JIANG Xiaohu,et al.Focal and global knowledge distillation for detectors[C]//Proceedings of 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).New Orleans:IEEE,2022:4633-4642.
[12] HE Kaiming,ZHANG Xiangyu,REN Shaoqing,et al.Deep residual learning for image recognition[C]//Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).Las Vegas:IEEE,2016:770-778.
[13] CHEN Guobin,CHOI W,YU Xiang,et al.Learning efficient object detection models with knowledge distillation[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems.Long Beach:ACM Press, 2017:742-751.
[14] SUN Ruoyu,TANG Fuhui,ZHANG Xiaopeng,et al.Distilling object detectors with task adaptive regularization[EB/OL].[2020-6-23].https://arxiv.org/abs/2006.13108.
[15] WANG Tao,YUAN Li,ZHANG Xiaopeng,et al.Distilling object detectors with fine-grained feature imitation[C]//Proceedings of 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Long Beach:IEEE,2019:4928-4937.
[16] CARION N,MASSA F,SYNNAEVE G,et al.End-to-end object detection with transformers[C]//Proceedings of 2020 European Conference on Computer Vision.Cham:Springer,2020:213-229.

针对浒苔目标检测的全局背景强化的位置蒸馏方法

Distillation method based on global background strengthen for Enteromorpha prolifera target detection

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]	高墨通, 杨维芳, 刘祖昱, 曹小双, 张瑞琪, 侯宇豪. 结合卷积神经网络和注意力机制的LSTM采空区地表沉降预测方法[J]. 测绘通报, 2024, 0(6): 53-58,170.
[2]	钟青岑, 吴晨昊, 向隆刚, 姚鹏. 顾及路网约束的深度地图匹配方法[J]. 测绘通报, 2024, 0(6): 96-102,133.
[3]	尹泽中, 李功权. 三维场叠加视频流的电子围栏越界检测方法[J]. 测绘通报, 2024, 0(6): 103-108.
[4]	王丽霞, 喜文飞, 史正涛, 赵子龙, 钱堂慧, 赵磊, 马奕捷. 深度学习滑坡识别算法中样本不平衡问题的研究[J]. 测绘通报, 2024, 0(5): 12-18.
[5]	姚楚羡, 蔡皓楠, 张远波, 唐可懿, 詹璐, 周宝定. 基于轻量化车载设备的道路病害检测方法[J]. 测绘通报, 2024, 0(5): 147-150.
[6]	张楠, 张蕴灵. 基于深度学习模型的复杂困难路网提取方法[J]. 测绘通报, 2024, 0(5): 155-159.
[7]	林泊锟, 丁勇, 李登华. 一种基于深度学习与图像局部特征提取的边坡异常监测技术[J]. 测绘通报, 2024, 0(4): 23-28.
[8]	庞敏. 基于深度学习的国产卫片图斑提取[J]. 测绘通报, 2024, 0(4): 124-128,134.
[9]	陈昶, 张岩, 李坤衡, 杨润泽, 张俊彬, 梁彦荣. 基于高分七号影像自动提取东北黑土区侵蚀沟的方法[J]. 测绘通报, 2024, 0(3): 1-7.
[10]	张兰兰, 王红雷. 基于深度学习的耕地非农化遥感监测与时空分析——以开阳县为例[J]. 测绘通报, 2024, 0(3): 13-18.
[11]	王鸿雁, 车向红, 徐辛超, 徐胜华, 李洪胜. 利用DeepLabv3+模型提取分析街景图像绿视率——以北京三环内为例[J]. 测绘通报, 2024, 0(3): 88-94.
[12]	马锦山, 贾国焕, 张赛, 张炯. 基于多源高分辨率遥感影像的典型自然资源要素提取[J]. 测绘通报, 2024, 0(3): 123-126,150.
[13]	王红雷, 严文璞. 基于深度学习技术的贵州省耕地占用遥感监测[J]. 测绘通报, 2024, 0(3): 134-139.
[14]	刘丽, 李士垚, 王润, 刘少宇, 宋永飞, 牛瑞卿. 联合深度学习与面向对象分析的卫宁北山露天矿山采场信息提取[J]. 测绘通报, 2024, 0(2): 51-57.
[15]	黄应华, 董振川, 李昊, 陈壮, 刘长睿, 张献州. 城市竣工测绘典型要素语义分割PointNet++深度学习模型适用性分析[J]. 测绘通报, 2024, 0(2): 85-89.