一种深度融合机制的遥感图像目标检测技术

doi:10.19665/j.issn1001-2400.2021.05.016

摘要/Abstract

摘要：

遥感图像覆盖幅面广、纹理信息丰富,其目标具有尺寸多样性,排列密集且与背景易混淆等特性,给快速定位和精准识别目标带来较多困难,尤其是易漏检小目标等。针对此问题,提出一种深度融合机制的遥感图像目标检测技术。该技术基于深度卷积神经网络,将多尺度、注意力机制与宽度学习三者融合,用于遥感图像目标检测技术。该技术首先基于多尺度与空间注意力机制获取到遥感图像的候选区域信息,然后采用通道注意力机制获取其多个尺度的特征信息并融合互补,旨在有效聚焦图像深层的高层语义信息和底层的小目标特征信息;最后,针对宽度学习存在超参数的确定需要依据不同的遥感图像,进行手工调参问题,提出基于贝叶斯网络搜索优化策略的宽度学习方法。该方法可智能地学习到一组适应于不同遥感图像数据集的超参数,对目标进行高效识别。实验结果证明,与当前先进的方法相比,该算法能够有效解决遥感图像目标检测中速度慢、精度低、易丢失小目标等问题,提升目标检测的准确率。

关键词: 深度卷积神经网络, 遥感图像, 注意力机制, 宽度学习, 目标检测

Abstract:

A new target detection technology for remote sensing images based on the deep fusion mechanism is proposed,which combines the multi-scale,attention mechanism and broad learning system based on the deep convolutional neural network.This technology focuses effectively on the high-level semantic information of remote sensing images and the characteristics of small targets.Because of the problem of manual adjustment of hyperparameters in the broad learning system,the author proposes a broad learning system based on the Bayesian network search,which can learn intelligently.A set of parameter values applicable to different remote sensing images can efficiently identify targets.Compared with other state-of-the-art methods,experimental results show that this technology can effectively solve the problems of a slow detection speed,a low recognition accuracy,and small targets in remote sensing image target detection tasks.

Key words: deep convolution neural network, remote sensing image, attention mechanism, broad learning system, object detection

中图分类号:

TP391.41

董如婵,焦李成,赵进,沈维燕. 一种深度融合机制的遥感图像目标检测技术[J]. 西安电子科技大学学报, 2021, 48(5): 128-138.

DONG Ruchan,JIAO Licheng,ZHAO Jin,SHEN Weiyan. Application of the deep fusion mechanism in object detection of remote sensing images[J]. Journal of Xidian University, 2021, 48(5): 128-138.

图/表 13

图1

图2

图3

表1

表2

表3

图4

图5

表4

图6

图7

表5

图8

参考文献 35

[1]	BOCHKOVSKIY A, WANG C Y, LIAO H. YOLOv4:Optimal Speed and Accuracy of Object Detection (2020)[J/OL]. [2020-04-23]. https://arxiv.org/pdf/2004.10934v1.pdf.
[2]	GIRSHICK R. Fast R-CNN[C]// Proceedings of the IEEE International Conference on Computer Vision.Piscataway:IEEE, 2019:510-519.
[3]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2014:580-587.
[4]	HE K, GKIOXARI G, DOLLAR P, et al. Mask R-CNN[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2):386-397. doi: 10.1109/TPAMI.34
[5]	LIU W, ANGUELOV D, ERHAN D, et al. SSD:Single Shot MultiBox Detector[C]// European Conference on Computer Vision.Heidelberg:Springer, 2016:21-37.
[6]	HE K, ZHANG X, REN S, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(9):1904-1916. doi: 10.1109/TPAMI.2015.2389824
[7]	REDMON J, DIVVALA S, GIRSHICK R, et al. You Only Look Once:Unified,Real-Time Object Detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2016:779-788.
[8]	REDMON J, FARHADI A. YOLO9000:Better,Faster,Stronger[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2017:6517-6525.
[9]	REDMON J, FARHADI A. YOLOv3:An Incrementalx Improvement (2018)[J/OL]. [2018-04-08]. https://arxiv.org/pdf/1804.02767.pdf.
[10]	REN S, HE K, 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. doi: 10.1109/TPAMI.2016.2577031
[11]	崔艳鹏, 王元皓, 胡建伟. 一种改进YOLOv3的动态小目标检测方法[J]. 西安电子科技大学学报, 2020, 47(3):1-7.
	CUI Yanpeng, WANG Yuanhao, HU Jianwei. Detection Method for a Dynamic Small Target Using the Improved YOLOv3[J]. Journal of Xidian University, 2020, 47(3):1-7.
[12]	汪昱东, 郭继昌, 王天保. 一种改进的雾天图像行人和车辆检测算法[J]. 西安电子科技大学学报, 2020, 47(4):70-77.
	WANG Yudong, GUO Jichang, WANG Tianbao. Algorithm for Foggy-Image Pedestrian and Vehicle Detection[J]. Journal of Xidian University, 2020, 47(4):70-77.
[13]	LIN T, GOYAL P, GIRSHICK R, et al. Focal Loss for Dense Object Detection[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2020, 42(2):318-327. doi: 10.1109/TPAMI.34
[14]	LONG Y, GONG Y, XIAO Z, et al. Accurate Object Localization in Remote Sensing Images Based on Convolutional Neural Networks[J]. IEEE Transactions on Geoscience Remote Sensing, 2017, 55(5):2486-2498. doi: 10.1109/TGRS.36
[15]	CHENG G, ZHOU P, HAN J. RIFD-CNN:Rotation-Invariant and Fisher Discriminative Convolutional Neural Networks for Object Detection[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2016:2884-2893.
[16]	CHENG G, HAN J, ZHOU P, et al. Multi-Class Geospatial Object Detection and Geographic Image Classification Based on Collection of Part Detectors[J]. ISPRS Journal of Photogrammetry Remote Sensing, 2014, 98:119-132. doi: 10.1016/j.isprsjprs.2014.10.002
[17]	CHENG G, HAN J. A Survey on Object Detection in Optical Remote Sensing Images[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 117:11-28. doi: 10.1016/j.isprsjprs.2016.03.014
[18]	DENG Z, SUN H, ZHOU S, et al. Multi-Scale Object Detection in Remote Sensing Imagery with Convolutional Neural Networks[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2018, 145:3-22. doi: 10.1016/j.isprsjprs.2018.04.003
[19]	GUO W, YANG W, ZHANG H, et al. Geospatial Object Detection in High Resolution Satellite Images Based on Multi-Scale Convolutional Neural Network[J]. Remote Sensing, 2018, 10:131. doi: 10.3390/rs10010131
[20]	WOO S, PARK J, LEE J, et al. CBAM:Convolutional Block Attention Module[C]// European Conference on Computer Vision.Heidelberg:Springer, 2018:3-19.
[21]	ZOU F, XIAO W, JI W, et al. Arbitrary-Oriented Object Detection via Dense Feature Fusion and Attention Model for Remote Sensing Super-Resolution Image[J]. Neural Computing Applications, 2020, 32(6):14549-14562. doi: 10.1007/s00521-020-04893-9
[22]	WANG C, BAI X, WANG S, et al. Multiscale Visual Attention Networks for Object Detection in VHR Remote Sensing Images[J]. IEEE Geoscience and Remote Sensing Letters, 2019, 16(2):310-314. doi: 10.1109/LGRS.2018.2872355
[23]	CHEN C L P, LIU Z. Broad Learning System:An Effective and Efficient Incremental Learning System without the Need for Deep Architecture[J]. IEEE Transactions on Neural Networks Learning Systems, 2018, 29(1):10-24. doi: 10.1109/TNNLS.5962385
[24]	KONG Y, WANG X, CHENG Y, et al. Hyperspectral Imagery Classification Based on Semi-Supervised Broad Learning System[J]. Remote Sensing, 2018, 10(5):685-692. doi: 10.3390/rs10050685
[25]	DONG R, XU D, JIAO L, et al. A Fast Deep Perception Network for Remote Sensing Scene Classification[J]. Remote Sensing, 2020, 12(4):729. doi: 10.3390/rs12040729
[26]	HE K, ZHANG X, REN S, et al. Deep Residual Learning for Image Recognition[C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2016:770-778.
[27]	ZAGORUYKO S, KOMODAKIS N. Paying More Attention to Attention:Improving the Performance of Convolutional Neural Networks via Attention Transfer(2016)[J/OL]. [2016-12-12]. https://arxiv.org/pdf/1612.03928.pdf.
[28]	ELSKEN T, METZEN J H, HUTTER F. Neural Architecture Search:A Survey (2018)[J/OL]. [2018-08-16]. https://arxiv.org/pdf/1808.05377.pdf.
[29]	CHEN C P, LIU Z. Broad Learning System:An Effective and Efficient Incremental Learning System without the Need for Deep Architecture[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(1):10-24. doi: 10.1109/TNNLS.5962385
[30]	SIGOP T. Bayesian Optimization Primer(2020)[EB/OL]. [2020-09-30]. https://sigopt.com/research.
[31]	ZOU Z, SHI Z. Random Access Memories:A New Paradigm for Target Detection in High Resolution Aerial Remote Sensing Images[J]. IEEE Transactions on Image Processing, 2018, 27(3):1100-1111. doi: 10.1109/TIP.2017.2773199
[32]	CHENG G, ZHOU P, HAN J. 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. doi: 10.1109/TGRS.2016.2601622
[33]	DONG R, XU D, ZHAO J, et al. Sig-NMS-Based Faster R-CNN Combining Transfer Learning for Small Target Detection in VHR Optical Remote Sensing Imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2019, 57(11):8534-8545. doi: 10.1109/TGRS.36
[34]	XU Z, XU X, WANG L, et al. Deformable ConvNet with Aspect Ratio Constrained NMS for Object Detection in Remote Sensing Imagery[J]. Remote Sensing, 2017, 9(12):1312-1330. doi: 10.3390/rs9121312
[35]	HAN X, ZHONG Y, ZHANG L. An Efficient and Robust Integrated Geospatial Object Detection Framework for High Spatial Resolution Remote Sensing Imagery[J]. Remote Sensing, 2017, 9(7):666-687. doi: 10.3390/rs9070666

数据集	飞机	油桶	轮船	合计
LEVIR	154	827	459	1 440
RSOD	1 503	2		1 505
NWPU VHR-10	7	91	5	201

类别	LARGE-RAM^[31]	Faster-RCNN^[33]	Sig-NMS^[33]	DMAB-Net
airplane	71.7	80.9	86.8	89.4
ship	60.8	79.0	79.4	84.9
oil-tank	43.0	60.6	68.2	68.6
mAP	58.5	73.5	78.1	80.8

类别	Faster-RCNN^[33]	Deformable^[34]	Sig-NMS^[33]	DMAB-Net
airplane	70.8	71.9	80.6	86.0
oil-tank	90.1	90.4	90.6	90.9
overpass	80.3	89.6	87.4	90.5
playground	98.1	99.7	99.1	100
mAP	84.8	87.9	89.4	91.7

类别	RICNN^[32]	LARGE- RAM^[31]	R-P-Faster R-CNN^[35]	Deformable^[34]	Sig-NMS^[33]	DMAB-Net
airplane	88.4	94.1	90.4	87.3	90.8	90.9
ship	77.3	85.5	75.0	81.4	80.5	87.8
oil-tank	85.3	85.9	44.4	63.6	59.2	58.5
Baseball diamond	88.1	89.9	89.9	90.4	90.8	99.6
Tennis court	40.8	67.1	79.0	81.6	80.8	88.5
basketball court	58.5	63.9	77.6	74.1	90.9	90.7
ground track field	86.7	48.9	87.7	90.3	99.8	100
harbor	68.6	62.9	79.1	75.3	90.3	98.2
bridge	61.5	58.7	68.2	71.4	67.8	73.5
vehicle	71.1	81.6	73.2	75.5	78.1	86.7
mAP	72.6	73.8	76.5	79.1	82.9	87.4