Remote Sensing Images Detection Based on Dense Connection and Feature Enhancement

doi:10.19678/j.issn.1000-3428.0061482

Abstract

Abstract: With the rapid development of remote sensing technology, remote sensing image detection is being widely used in military, agriculture, transportation, urban planning, and other fields.With the continuous improvements of the remote sensing resolution and data volume, a manual data processing method has been unable to meet the real-time requirements.Therefore, the realization of an efficient and accurate automatic data processing approach has become a research hotspot in this field.Aiming at the characteristics of a high resolution, complex background, and small target scale of a remote sensing image, this paper proposes an improved YOLOv3 algorithm to improve the detection effect of such an image.Based on the original YOLOv3 algorithm, the improved dense connection network replaces the original DarkNet53 as the basic network, thereby improving the scale diversity of network input and prediction results.This study reduces the input loss through the valve stem module, and adds a feature enhancement module to the shallow feature map, thereby enriching the receptive field of the feature map and strengthening the extraction of the shallow feature information of the network, while ensuring the overall detection performance, this study improves the detection accuracy and robustness of the network for medium and small targets in remote sensing images.This study also carried out several groups of comparative experiments on remote sensing image datasets.The results show that, compared with the original YOLOv3 algorithm, the average accuracy of the algorithm is improved by 9.45 percentage points, and the detection accuracy of small-scale targets is increased even more significantly, reaching 11.03 percentage points.In addition, the number of model parameters is effectively reduced.

Key words: target detection, deep learning, remote sensing image, feature enhancement, dense connection

摘要： 遥感技术的快速发展使得遥感图像检测技术广泛应用于军事、农业、交通、城市规划等多个领域。随着遥感分辨率和数据体量的不断提升，通过人工处理数据的方法已经无法满足实时性需求，因此，实现高效、精准的自动化数据处理方式成为该领域的研究热点。针对遥感图像分辨率高、背景复杂、目标尺度小等特点，提出一种改进的YOLOv3算法，用以提升遥感图像的检测效果。在原始YOLOv3算法的基础上，使用改进的密集连接网络替换原有的DarkNet53作为基础网络，以提升网络输入和预测结果的尺度多样性。通过阀杆模块降低输入损失，同时在浅层特征图中加入特征增强模块，从而丰富特征图的感受野，强化网络对浅层特征信息的提取，在保证整体检测性能的同时使网络对遥感图像中、小目标的检测精度和鲁棒性均有所提升。在遥感图像数据集上进行多组对比实验，结果表明，相比原始YOLOv3算法，该算法的平均准确率提高9.45个百分点，在小尺度目标上的检测准确率提升更显著，达到11.03个百分点，且模型参数量得到有效缩减。

关键词: 目标检测, 深度学习, 遥感图像, 特征增强, 密集连接

CLC Number:

TP391

WANG Daolei, DU Wenbin, LIU Yiteng, ZHANG Tianyu, SUN Jiajun, LI Mingshan. Remote Sensing Images Detection Based on Dense Connection and Feature Enhancement[J]. Computer Engineering, 2022, 48(6): 251-256,262.

王道累, 杜文斌, 刘易腾, 张天宇, 孙嘉珺, 李明山. 基于密集连接与特征增强的遥感图像检测[J]. 计算机工程, 2022, 48(6): 251-256,262.

/ / Recommend / Download Citations

URL: http://www.ecice06.com/EN/10.19678/j.issn.1000-3428.0061482

http://www.ecice06.com/EN/Y2022/V48/I6/251

Figures/Tables 11

References

[1] WU Y R, HU H, ZHANG Y.Multi-resolution object detection and data fusion for large-scale remote sensing images based on deep learning method[C]//Proceedings of IEEE Advanced Information Management, Communicates, Electronic and Automation Control Conference.Washington D.C., USA:IEEE Press, 2019:933-937.
[2] FELZENSZWALB P F, GIRSHICK R B, MCALLESTER D, et al.Object detection with discriminatively trained part-based models[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2010, 32(9):1627-1645.
[3] 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.
[4] HE K M, GKIOXARI G, DOLLÁR P, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:2980-2988.
[5] LIU W, ANGUELOV D, ERHAN D, et al.SSD:Single Shot MultiBox Detector[M].Berlin, Germany:Springer, 2016.
[6] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:779-788.
[7] REDMON J, FARHADI A.YOLO9000:better, faster, stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:6517-6525.
[8] REDMON J, FARHADI A.YOLOv3:an incremental improvement[EB/OL].[2021-03-05].https://arxiv.org/pdf/1804.02767.pdf.
[9] ZHENG Z, LEI L, SUN H, et al.A review of remote sensing image object detection algorithms based on deep learning[C]//Proceedings of IEEE International Conference on Image, Vision and Computing.Washington D.C., USA:IEEE Press, 2020:34-43.
[10] SZEGEDY C, IOFFE S, VANHOUCKE V, et al.Inception-v4, inception-ResNet and the impact of residual connections on learning[EB/OL].[2021-03-05].https://arxiv.org/pdf/1602.07261.pdf.
[11] LIU S, HUANG D.Receptive field block net for accurate and fast object detection[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2018:385-400.
[12] LIN T Y, DOLLÁR P, GIRSHICK R, et al.Feature pyramid networks for object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:936-944.
[13] 李昕, 赵猛, 董修武, 等.基于改进YOLOV3算法的遥感图像油罐检测[J].中国科技论文, 2020, 15(3):267-273. LI X, ZHAO M, DONG X W, et al.Oil tank detection in remote sensing images based on improved YOLOV3 algorithm[J].China Sciencepaper, 2020, 15(3):267-273.(in Chinese)
[14] 郑海生, 王雪纯.面向遥感图像实时目标检测任务的轻量级神经网络设计与实现[J].电脑与电信, 2020(3):18-22. ZHENG H S, WANG X C.Design and implementation of lightweight neural network for real-time target detection tasks in remote-sensing images[J].Computer & Telecom-munication, 2020(3):18-22.(in Chinese)
[15] 沈丰毅, 王琴, 蒋剑飞.基于改进卷积神经网络舰船的实时目标检测算法[J].计算机应用研究, 2020, 37(S1):316-319. SHEN F Y, WANG Q, JIANG J F.Real-time automatic ship detection based on modified convolutional neural networks[J].Application Research of Computers, 2020, 37(S1):316-319.(in Chinese)
[16] HUANG G, LIU Z, VAN DER MAATEN L, et al.Densely connected convolutional networks[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:2261-2269.
[17] SHEN Z Q, LIU Z, LI J G, et al.DSOD:learning deeply supervised object detectors from scratch[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:1937-1945.
[18] HE K, ZHANG X, REN S, et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[EB/OL].[2021-03-05].https://ieeexplore.ieee.org/document/7005506/.
[19] CHENG G, ZHOU P C, HAN J W.Learning rotation-invariant convolutional neural networks for object detection in VHR optical remote sensing images[J].IEEE Transactions on Geoscience and Remote Sensing, 2016, 54(12):7405-7415.
[20] TAN M X, PANG R M, LE Q V.EfficientDet:scalable and efficient object detection[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2020:10778-10787.
[21] DUAN K W, BAI S, XIE L X, et al.CenterNet:keypoint triplets for object detection[C]//Proceedings of IEEE/CVF International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2019:6568-6577.
[22] BOCHKOVSKIY A, WANG C Y, LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2021-03-05].https://arxiv.org/pdf/2004.10934.pdf.

Please choose a citation manager

Content to export