Research on Object Detection Algorithm for Remote Sensing Images Based on Multi-Scale Feature Fusion

doi:10.3778/j.issn.1002-8331.2403-0239

Abstract

Abstract: In order to tackle the challenges of complex background, feature conflict and variable target scale in remote sensing images, a target detection method based on multi-scale feature fusion using YOLOv8s as the baseline model is proposed for remote sensing images. Firstly, RepVGG network is used as the feature extraction network to improve the feature extraction capability and effectively achieve the capture of global semantic information. Secondly, a tripolar integrative fusion (TIF) module is designed in the neck network to improve the detection accuracy of targets at all scales through the effective fusion of positional and semantic information. Finally, the SlideLoss function is used as the classification loss function to enhance the detection ability of difficult targets and enhance the precision of target detection. The experimental results indicate that the improved model achieves 94.4%, 93.0% and 95.5% detection accuracies on the NWPU VHR-10, RSOD and UCAS-AOD datasets respectively, which are 5.1, 6.0 and 4.4?percentape points higher than the baseline model, which is superior to the other methods in terms of accuracy and is able to better complete target detection task in remote sensing images.

Key words: deep learning, object detection, remote sensing image, multi-scale feature fusion

摘要： 针对遥感图像背景复杂、特征冲突以及目标尺度多变等问题，以YOLOv8s为基线模型，提出一种基于多尺度特征融合的遥感图像目标检测方法。将RepVGG网络作为模型的特征提取网络，提高模型的特征提取能力，有效实现全局语义信息的捕获；在颈部网络提出了一种三极综合性融合模块（tripolar integrative fusion，TIF），通过位置信息和语义信息的有效融合，提高对各个尺度目标的检测精度；将SlideLoss滑动损失函数作为模型的分类损失函数，增强模型对困难目标的检测能力，从而提高目标检测的准确性。实验结果表明，改进后的模型在NWPU VHR-10、RSOD和UCAS-AOD数据集上分别取得了94.4%、93.0%和95.5%的检测精度，相较基线模型分别提升了5.1、6.0、4.4个百分点，在准确性方面优于其他方法，能更好地完成遥感图像目标检测任务。

关键词: 深度学习, 目标检测, 遥感图像, 多尺度特征融合

XU Siyuan, WU Weilin. Research on Object Detection Algorithm for Remote Sensing Images Based on Multi-Scale Feature Fusion[J]. Computer Engineering and Applications, 2024, 60(23): 249-256.

许思源, 吴伟林. 多尺度特征融合的遥感图像目标检测算法研究[J]. 计算机工程与应用, 2024, 60(23): 249-256.

References

[1] 王子健, 王云艳, 武华轩. 基于改进YOLOX的遥感图像目标检测[J]. 扬州大学学报 (自然科学版), 2023, 26(5): 64-71.
WANG Z J, WANG Y Y, WU H X. Remote sensing image target detection based on improved YOLOX[J]. Journal of Yangzhou University (Natural Science Edition), 2023, 26(5): 64-71.
[2] CHENG G, YUAN X, YAO X W, et al. Towards large-scale small object detection: survey and benchmarks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023, 45(11): 13467-13488.
[3] 胡俊, 顾晶晶, 王秋红. 基于遥感图像的多模态小目标检测[J]. 图学学报, 2022, 43(2): 197-204.
HU J, GU J J, WANG Q H. Multimodal small target detection based on remote sensing image[J]. Journal of Graphics, 2022, 43(2): 197-204.
[4] WANG C S, WANG Q, WU H R, et al. Low-altitude remote sensing opium poppy image detection based on modified YOLOv3[J]. Remote Sensing, 2021, 13(11): 2130.
[5] GOLDMAN E, HERZIG R, EISENSCHTAT A, et al. Precise detection in densely packed scenes[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019: 5222-5231.
[6] SHI F B, LIU Y, WANG H Y. Target detection in remote sensing images based on multi-scale fusion faster rcnn[C]//Proceedings of the 35th Chinese Control and Decision Conference, 2023: 4043-4046.
[7] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierar-chies for accurate object detection and semantic segmentation[C]//Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
[8] TAN M X, LE Q. EfficientNet: rethinking model scaling for convolutional neural networks[C]//Proceedings of the International Conference on Machine Learning, 2019: 6105-6114.
[9] 赵加坤, 孙俊, 韩睿, 等. 基于改进的Faster Rcnn遥感图像目标检测[J]. 计算机应用与软件, 2022, 39(5): 192-196.
ZHAO J K, SUN J, HAN R, et al. Object detection based on improved Faster Rcnn for remote sensing image[J]. Computer Applications and Software, 2022, 39(5): 192-196.
[10] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149.
[11] ZHOU L M, LI Y H, RAO X H, et al. Feature enhancement-based ship target detection method in optical remote sensing images[J]. Electronics, 2022, 11(4): 634.
[12] BOCHKOVSKIY A, WANG C Y, LIAO H Y M. YOLOv4: optimal speed and accuracy of object detection[J]. arXiv:2004. 10934, 2020.
[13] XU X K, FENG Z J, CAO C Q, et al. An improved swin transformer-based model for remote sensing object detection and instance segmentation[J]. Remote Sensing, 2021, 13(23): 4779.
[14] LIU Z, LIN Y T, CAO Y, et al. Swin transformer: hierarchical vision transformer using shifted windows[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision, 2021: 10012-10022.
[15] DING X H, ZHANG X Y, MA N N, et al. RepVGG: making VGG-style convnets great again[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 13733-13742.
[16] YU Z P, HUANG H B, CHEN W J, et al. YOLO-facev2: a scale and occlusion aware face detector[J]. arXiv:2208.02019, 2022.
[17] TERVEN J, CORDOVA-ESPARZA D. A comprehensive review of YOLO: from YOLOv1 to YOLOv8 and beyond[J]. arXiv:2304.00501, 2023.
[18] SIMONYAN K, ZISSERMANA. Very deep convolutional networks for large-scale image recognition[J]. arXiv:1409. 1556, 2014.
[19] HAN K, WANG Y H, TIAN Q, et al. GhostNet: more features from cheap operations[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2020: 1580-1589.
[20] CHENG G, HAN J W, ZHOU P C, et al. Multi-class geospatial object detection and geographic image classification based on collection of part detectors[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014, 98: 119-132.
[21] LONG Y, GONG Y P, XIAO Z F, et al. Accurate object localization in remote sensing images based on convolutional neural networks[J]. IEEE Transactions on Geoscience and Remote Sensing, 2017, 55: 2486-2498.
[22] ZHU H G, CHEN X G, DAI W Q, et al. Orientation robust object detection in aerial images using deep convolutional neural network[C]//Proceedings of the 2015 IEEE International Conference on Image Processing, 2015: 3735-3739.
[23] CHEN J R, KAO S H, HE H, et al. Run, don’t walk: chasing higher flops for faster neural networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 12021-12031.
[24] MA N N, ZHANG X Y, ZHENG H T, et al. Shufflenet v2: practical guidelines for efficient CNN architecture design[C]//Proceedings of the European Conference on Computer Vision, 2018: 116-131.
[25] CHEN H T, WANG Y H, GUO J Y, et al. VanillaNet: the power of minimalism in deep learning[J]. arXiv:2305. 12972, 2023.
[26] ZHU L, WANG X J, KE Z H, et al. BiFormer: vision transformer with bi-level routing attention[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023: 10323-10333.
[27] CUI C, GAO T Q, WEI S Y, et al. PP-LCNet: a lightweight cpu convolutional neural network[J]. arXiv:2109.15099, 2021.
[28] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the IEEE International Conference on Computer Vision, 2017: 2980-2988.
[29] ZHANG H Y, WANG Y, DAYOUB F, et al. VarifocalNet: an IoU-aware dense object detector[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2021: 8514-8523.
[30] TANG S Y, FANG Y N, ZHANG S. HIC-YOLOv5: improved YOLOv5 for small object detection[J]. arXiv:2309.16393, 2023.
[31] ZHANG Z X. Drone?YOLO: an efficient neural network method for target detection in drone images[J]. Drones, 2023, 7(8): 526.
[32] GE Z, LIU S T, WANG F, et al. YOLOX: exceeding YOLO series in 2021[J]. arXiv:2107.08430, 2021.
[33] LV W Y, XU S L, ZHAO Y, et al. DETRs beat YOLOs on real-time object detection[J]. arXiv:2304.08069, 2023.