基于双分类头的遥感图像精细化目标检测方法

张锋; 滕书华; 韩幸; 王应谦; 王雪莹

doi:10.11834/jrs.20254243

模型与方法 | 浏览量 : 0 下载量: 338 CSCD: 0

R-PDF
PDF
导出
分享
收藏
专辑

基于双分类头的遥感图像精细化目标检测方法
A fine-grained obect detection method for remote sensing images based on dual classification head
2026年30卷第1期页码：170-182
收稿：2024-06-14，

纸质出版：2026-01-07
DOI： 10.11834/jrs.20254243
稿件说明：

移动端阅览

张锋，滕书华，韩幸，王应谦，王雪莹.2026.基于双分类头的遥感图像精细化目标检测方法.遥感学报，30（1）： 170-182 DOI： 10.11834/jrs.20254243.

Zhang F，Teng S H，Han X，Wang Y Q and Wang X Y. 2026. A fine-grained obect detection method for remote sensing images based on dual classification head. National Remote Sensing Bulletin， 30（1）：170-182 DOI： 10.11834/jrs.20254243.

摘要

高分辨率遥感图像的可获得性大幅提升，使得遥感图像目标精细化检测成为了遥感以及计算机视觉领域重要的研究方向。针对遥感图像目标精细化检测中存在的相似数据利用不充分、错误标签影响模型精度和相似类别难以区分的问题，本文提出了一种基于双分类头的遥感图像精细化目标检测方法。首先，针对遥感图像精细化目标检测中无法有效利用相似数据的问题，提出一种双分类检测头，不同的分类头分别对不同数据集训练，让类别定义不同的相似数据共同参与训练，进而有效利用相似数据，显著提升了模型精度。其次，针对训练标签噪声问题，设计一种基于预测的错误标签过滤方法，减小错误标签对模型训练的影响。最后，针对精细化目标检测中类内差异大、类间差异小的问题，定义一种Margin交叉熵损失，通过增大分类边界提升了模型精度。在精细化遥感目标检测竞赛数据集和FAIR1M数据集上的实验表明，本文提出的方法显著提高了遥感影像目标精细化检测的精度和鲁棒性。代码已开源在

https：//github.com/zf020114/DCH

https://github.com/zf020114/DCH

。

Abstract

With the availability of remote sensing images improved

fine-grained object detection in remote sensing images has become an important research topic in the field of remote sensing and computer vision

and a large number of remote sensing image object detection datasets have been published. However

the current fine-grained object detection method cannot effectively use these publicized datasets

although they have similar image scenes and similar categories with the fine-grained object detection datasets

which will greatly affect the accuracy of the detectors. To solve the problem of insufficient utilization of similar data

incorrect labels affecting model accuracy

and difficulty in distinguishing similar categories

a fine-grained object detection method based on a dual classification head is proposed in this paper. Firstly

to solve the problem of similar data not being able to be used effectively in fine-grained object detection in remote sensing images

a dual classified detection head is proposed in this paper. This head uses different classified branches to train different datasets and allow similar data with varying definitions of category to participate in the training

which can effectively use similar data and significantly improve the accuracy of th

e model. Secondly

to solve the problem of noise labels in training data in fine-grained object detection

an error label filtering module based on prediction is proposed to reduce the impact of error labels. The category and confidence score of the object were obtained by an additional SoftMax operation in training. If the predicted category is different from the ground truth label and the confidence score is higher than a certain threshold

the label is considered to be incorrect. Finally

to solve the problem of large intra-class distance and small inter-class distance in fine-grained object detection in high-resolution remote sensing images

a margin cross-entropy loss function is designed to increase the classification boundary. The detection accuracy was improved by increasing the classification boundary of different categories. The Margin cross-entropy loss function calculates the loss by artificially subtracting the constant term from the network-predicted value of the positive sample and the detection accuracy was improved by increasing the classification boundary of different categories. Experiments on remote sensing fine-grained object detection datasets and FAIR1M datasets show that the proposed method improves the accuracy and robustness of fine-grained object detectors significantly in remote sensing images. The code is open source at

https://github.com/zf020114/DCH

关键词

Keywords

references

Deng J K , Guo J , Xue N N and Zafeiriou S . 2019 . ArcFace: additive angular margin loss for deep face recognition // Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Long Beach : IEEE: 4685 - 4694 [ DOI: 10.1109/CVPR.2019.00482 http://dx.doi.org/10.1109/CVPR.2019.00482 ]

Ding J , Xue N , Long Y , Xia G S and Lu Q K . 2018 . Learning RoI transformer for oriented object detection in aerial images // Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . Long Beach : IEEE: 2844 - 2853 [ DOI: 10.1109/CVPR.2019.00296 http://dx.doi.org/10.1109/CVPR.2019.00296 ]

Han B , Yao Q M , Yu X R , Niu G , Xu M , Hu W H , Tsang I W and Sugiyama M . 2018 . Co-teaching: robust training of deep neural networks with extremely noisy labels // Proceedings of the Proceedings of the 32nd International Conference on Neural Information Processing Systems . Montréal : Curran Associates Inc.: 8536 - 8546 [ DOI: 10.5555/3327757.3327944 http://dx.doi.org/10.5555/3327757.3327944 ]

Han J M , Ding J , Xue N and Xia G S . 2021a . ReDet: a rotation-equivariant detector for aerial object detection // Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Nashville : IEEE: 2785 - 2794 [ DOI: 10.1109/CVPR46437.2021.00281 http://dx.doi.org/10.1109/CVPR46437.2021.00281 ]

Han J M , Ding J , Li J and Xia G S . 2021b . Align deep features for oriented object detection . IEEE Transactions on Geoscience and Remote Sensing , 60 : 5602511 [ DOI: 10.1109/TGRS.2021.3062048 http://dx.doi.org/10.1109/TGRS.2021.3062048 ]

Han Y Q , Yang X Y , Pu T and Peng Z M . 2022 . Fine-grained recognition for oriented ship against complex scenes in optical remote sensing images . IEEE Transactions on Geoscience and Remote Sensing , 60 : 5612318 [ DOI: 10.1109/TGRS.2021.3123666 http://dx.doi.org/10.1109/TGRS.2021.3123666 ]

Hu Z B , Gao K , Zhang X D , Wang J W , Wang H and Han J W . 2022 . Probability differential-based class label noise purification for object detection in aerial images . IEEE Geoscience and Remote Sensing Letters , 19 : 6509705 [ DOI: 10.1109/LGRS.2022.3172983 http://dx.doi.org/10.1109/LGRS.2022.3172983 ]

Jiang Y Y , Zhu X Y , Wang X B , Yang S L , Li W , Wang H , Fu P and Luo Z B . 2018 . R 2 CNN: rotational region CNN for arbitrarily-oriented scene text detection // Proceedings of the 24th International Conference on Pattern Recognition . Beijing : IEEE: 3610 - 3615 [ DOI: 10.1109/ICPR.2018.8545598 http://dx.doi.org/10.1109/ICPR.2018.8545598 ]

Li D R , Wang M , Shen X and Dong Z P . 2017 . From earth observation satellite to earth observation brain . Geomatics and Information Science of Wuhan University , 42 ( 2 ): 143 - 149

李德仁 , 王密 , 沈欣 , 董志鹏 . 2017 . 从对地观测卫星到对地观测脑 . 武汉大学学报(信息科学版) , 42 ( 2 ): 143 - 149 [ DOI: 10.13203/j.whugis20160526 http://dx.doi.org/10.13203/j.whugis20160526 ]

Li K , Wan G , Cheng G , Meng L Q and Han J W . 2020 . Object detection in optical remote sensing images: a survey and a new benchmark . ISPRS Journal of Photogrammetry and Remote Sensing , 159 : 296 - 307 [ DOI: 10.1016/j.isprsjprs.2019.11.023 http://dx.doi.org/10.1016/j.isprsjprs.2019.11.023 ]

Li W T , Chen Y J , Hu K X and Zhu J K . 2022 . Oriented RepPoints for aerial object detection // Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition . New Orleans : IEEE: 1819 - 1828 [ DOI: 10.1109/CVPR52688.2022.00187 http://dx.doi.org/10.1109/CVPR52688.2022.00187 ]

Li Y X , Hou Q B , Zheng Z H , Chen M M , Yang J and Li X . 2023 . Large selective kernel network for remote sensing object detection // Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision (ICCV) . Paris : IEEE: 16748 - 16759 [ DOI: 10.1109/ICCV51070.2023.01540 http://dx.doi.org/10.1109/ICCV51070.2023.01540 ]

Luo H L and Chen H K . 2020 . Survey of object detection based on deep learning . Acta Electronica Sinica , 48 ( 6 ): 1230 - 1239

罗会兰 , 陈鸿坤 . 2020 . 基于深度学习的目标检测研究综述 . 电子学报 , 48 ( 6 ): 1230 - 1239 [ DOI: 10.3969/j.issn.0372-2112.2020.06.026 http://dx.doi.org/10.3969/j.issn.0372-2112.2020.06.026 ]

Ma J Q , Shao W Y , Ye H , Wang L , Wang H , Zheng Y B and Xue X Y . 2018 . Arbitrary-oriented scene text detection via rotation proposals . IEEE Transactions on Multimedia , 20 ( 11 ): 3111 - 3122 [ DOI: 10.1109/TMM.2018.2818020 http://dx.doi.org/10.1109/TMM.2018.2818020 ]

Pu Y F , Wang Y R , Xia Z F , Han Y Z , Wang Y L , Gan W H , Wang Z D , Song S J and Huang G . 2023 . Adaptive rotated convolution for rotated object detection // Proceedings of the 2023 IEEE/CVF International Conference on Computer Vision (ICCV) . Paris : IEEE: 6566 - 6577 [ DOI: 10.1109/ICCV51070.2023.00606 http://dx.doi.org/10.1109/ICCV51070.2023.00606 ]

Ren S Q , He K M , Girshick R and Sun J . 2015 . Faster R-CNN: towards real-time object detection with region proposal networks // Proceedings of the 29th International Conference on Neural Information Processing Systems . Montreal : MIT Press: 91 - 99 [ DOI: 10.5555/2969239.2969250 http://dx.doi.org/10.5555/2969239.2969250 ]

Sun X , Wang P J , Yan Z Y , Xu F , Wang R P , Diao W H , Chen J , Li J H , Feng Y C , Xu T , Weinmann M , Hinz S , Wang C and Fu K . 2022 . FAIR 1 m: a benchmark dataset for fine-grained object recognition in high-resolution remote sensing imagery. ISPRS Journal of Photogrammetry and Remote Sensing , 184 : 116 - 130 [ DOI: 10.1016/j.isprsjprs.2021.12.004 http://dx.doi.org/10.1016/j.isprsjprs.2021.12.004 ]

Wang Y S , Ma X J , Chen Z Y , Luo Y , Yi J F and Bailey J . 2019 . Symmetric cross entropy for robust learning with noisy labels // Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision (ICCV) . Seoul : IEEE: 322 - 330 [ DOI: 10.1109/ICCV.2019.00041 http://dx.doi.org/10.1109/ICCV.2019.00041 ]

Xia G S , Bai X , Ding J , Zhu Z , Belongie S , Luo J B , Datcu M , Pelillo M and Zhang L P . 2018 . DOTA: a large-scale dataset for object detection in aerial images // Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition . Salt Lake City : IEEE: 3974 - 3983 [ DOI: 10.1109/CVPR.2018.00418 http://dx.doi.org/10.1109/CVPR.2018.00418 ]

Xiao J S , Zhang S H , Chen Y H , Wang Y F and Yang L H . 2022 . Remote sensing image object detection based on bidirectional feature fusion and feature selection . Acta Electronica Sinica , 50 ( 2 ): 267 - 272

肖进胜 , 张舒豪 , 陈云华 , 王元方 , 杨力衡 . 2022 . 双向特征融合与特征选择的遥感影像目标检测 . 电子学报 , 50 ( 2 ): 267 - 272 [ DOI: 10.12263/DZXB.20210354 http://dx.doi.org/10.12263/DZXB.20210354 ]

Xie X X , Cheng G , Wang J B , Yao X W and Han J W . 2021 . Oriented R-CNN for object detection // Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision . Montreal : IEEE: 3500 - 3509 [ DOI: 10.1109/ICCV48922.2021.00350 http://dx.doi.org/10.1109/ICCV48922.2021.00350 ]

Xu Y C , Fu M T , Wang Q M , Wang Y K , Chen K , Xia G S and Bai X . 2021 . Gliding vertex on the horizontal bounding box for multi-oriented object detection . IEEE Transactions on Pattern Analysis and Machine Intelligence , 43 ( 4 ): 1452 - 1459 [ DOI: 10.1109/TPAMI.2020.2974745 http://dx.doi.org/10.1109/TPAMI.2020.2974745 ]

Yang X , Yan J C , Feng Z M and He T . 2019 . R 3 Det: refined single-stage detector with feature refinement for rotating object // Proceedings of the 39th AAAI Conference on Artificial Intelligence . Washington : AAAI Press: 3163 - 3171 [ DOI: 10.1609/aaai.v35i4.16426 http://dx.doi.org/10.1609/aaai.v35i4.16426 ]

Yao Y Q , Cheng G , Xie X X and Han J W . 2021 . Optical remote sensing image object detection based on multi-resolution feature fusion . National Remote Sensing Bulletin , 25 ( 5 ): 1124 - 1137

姚艳清 , 程塨 , 谢星星 , 韩军伟 . 2021 . 多分辨率特征融合的光学遥感图像目标检测 . 遥感学报 , 25 ( 5 ): 1124 - 1137 [ DOI: 10.11834/jrs.20210505 http://dx.doi.org/10.11834/jrs.20210505 ]

Yi J R , Wu P X , Liu B , Huang Q Y , Qu H and Metaxas D . 2021 . Oriented object detection in aerial images with box boundary-aware vectors // Proceedings of the 2021 IEEE Winter Conference on Applications of Computer Vision . Waikoloa : IEEE: 2149 - 2158 [ DOI: 10.1109/WACV48630.2021.00220 http://dx.doi.org/10.1109/WACV48630.2021.00220 ]

Yu H T , Tian Y J , Ye Q X and Liu Y F . 2024 . Spatial transform decoupling for oriented object detection // Proceedings of the 39th AAAI Conference on Artificial Intelligence . Vancouver : AAAI Press: 6782 - 6790 [ DOI: 10.1609/aaai.v38i7.28502 http://dx.doi.org/10.1609/aaai.v38i7.28502 ]

Yu X R , Han B , Yao J C , Niu G , Tsang I and Sugiyama M . 2019 . How does disagreement help generalization against label corruption,Proceedings of the 36th International Conference on Machine Learning . Long Beach : PMLR : 7164 - 7173 10.48550/arXiv.1901.04215 http://dx.doi.org/10.48550/arXiv.1901.04215

Zhang F , Wang X Y , Zhou S L and Wang Y Q . 2021 . DARDet: a dense anchor-free rotated object detector in aerial images . IEEE Geoscience and Remote Sensing Letters , 19 : 8024305 [ DOI: 10.1109/LGRS.2021.3122924 http://dx.doi.org/10.1109/LGRS.2021.3122924 ]

Zhang H , Li F , Liu S L , Zhang L , Su H , Zhu J , Ni L M and Shum H Y . 2023 . DINO: DETR with improved denoising anchor boxes for end-to-end object detection Proceedings of the International Conference on Learning Representations(ICLR) Kigali City [DOI: 10.48550/arXiv.2203.03605] Zhang L, Zhang Y S, Yu Y, Ma Y Z and Jiang H G . 2022 . Survey on object detection in tilting box for remote sensing images. National Remote Sensing Bulletin, 26 ( 9 ): 1723 - 1743

张磊 , 张永生 , 于英 , 马永政 , 姜怀刚 . 2022 . 遥感图像倾斜边界框目标检测研究进展与展望 . 遥感学报 , 26 ( 9 ): 1723 - 1743 [ DOI: 10.11834/jrs.20210247 http://dx.doi.org/10.11834/jrs.20210247 ]

Zhang S , Li S S , Wei G F , Zhang X N and Gao J W . 2022 . Refined multi-scale feature-oriented object detection of the remote sensing images . National Remote Sensing Bulletin , 26 ( 12 ): 2616 - 2628

张省 , 李山山 , 魏国芳 , 张新耐 , 高建威 . 2022 . 面向精细化多尺度特征的遥感图像目标检测 . 遥感学报 , 26 ( 12 ): 2616 - 2628 [ DOI: 10.11834/jrs.20221801 http://dx.doi.org/10.11834/jrs.20221801 ]

Zhang Z L and Sabuncu M R . 2018 . Generalized cross entropy loss for training deep neural networks with noisy labels // Proceedings of the 32nd International Conference on Neural Information Processing Systems . Montréal : Curran Associates Inc.: 8792 - 8802 [ DOI: 10.5555/3327546.3327555 http://dx.doi.org/10.5555/3327546.3327555 ]

Zhao P B , Qu Z S , Bu Y J , Tan W M and Guan Q Y . 2021 . PolarDet: a fast, more precise detector for rotated target in aerial images . International Journal of Remote Sensing , 42 ( 15 ): 5831 - 5861 [ DOI: 10.1080/01431161.2021.1931535 http://dx.doi.org/10.1080/01431161.2021.1931535 ]

Zhao Z Y , Xue Q L , He Y H , Bai Y F , Wei X and Gong Y H . 2024 . Projecting points to axes: oriented object detection via point-axis representation // Proceedings of the 18th European Conference on Computer Vision–ECCV 2024 . Milan : Springer: 161 - 179 [ DOI: 10.1007/978-3-031-73390-1_10 http://dx.doi.org/10.1007/978-3-031-73390-1_10 ]

Zhou X Y , Wang D Q and Krähenbühl P . 2019 . Objects as points . arXiv preprint arXiv : 1904 . 07850 [ doi: 10.48550/arXiv.1904.07850 http://dx.doi.org/10.48550/arXiv.1904.07850 ]