机载广域遥感图像的尺度归一化目标检测方法

doi:10.13229/j.cnki.jdxbgxb.20230034

吉林大学学报(工学版) ›› 2024, Vol. 54 ›› Issue (8): 2329-2337.doi: 10.13229/j.cnki.jdxbgxb.20230034

机载广域遥感图像的尺度归一化目标检测方法

朱圣杰^1,²(),王宣¹(),徐芳¹,彭佳琦³,王远超⁴

^1.中国科学院长春光学精密机械与物理研究所, 长春 130033
^2.中国科学院大学大珩学院, 北京 100049
^3.驻长春地区第一军事代表室, 长春 130033
^4.上海机电工程研究所, 上海 201109

收稿日期:2023-01-10 出版日期:2024-08-01 发布日期:2024-08-30
通讯作者: 王宣 E-mail:shengjie_zhu@foxmail.com;ally637@163.com
作者简介:朱圣杰（1997-），男，博士研究生. 研究方向：航空航天成像，计算机视觉. E-mail： shengjie_zhu@foxmail.com
基金资助:
国家自然科学基金项目(61905240)

Multi-scale normalized detection method for airborne wide-area remote sensing images

Sheng-jie ZHU^1,²(),Xuan WANG¹(),Fang XU¹,Jia-qi PENG³,Yuan-chao WANG⁴

^1.Changchun Institute of Optics，Fine Mechanics and Physics，Chinese Academy of Sciences，Changchun 130033，China
^2.Daheng College，University of Chinese Academy of Sciences，Beijing 100049，China
^3.First Military Representative Office in Changchun，Changchun 130033，China
^4.Shanghai Electro-Mechanical Engineering Institute，Shanghai 201109，China

Received:2023-01-10 Online:2024-08-01 Published:2024-08-30
Contact: Xuan WANG E-mail:shengjie_zhu@foxmail.com;ally637@163.com

摘要/Abstract

摘要：

针对机载广域遥感图像的目标尺寸变化大、背景噪声复杂以及局部目标密集给目标检测任务带来的困难，本文通过优化分割方法统一输入图像的目标像素尺寸，并以此简化模型结构提出了一种尺度归一化卷积神经网络模型MNNet。为增强局部之间的特征关联，本文设计了全局连接块（SGC），有效提高了检测的精度。针对现有非极大值抑制算法的超参数依赖经验设置的问题，本文提出了一种自适应非极大值抑制方法（DNMS），降低了模型的部署难度。在RSF数据集上的测试结果表明：本文模型的检测平均精度（AP）高于其他模型5.0%以上，在检测速度上达到了57.7 帧/s，可以满足遥感图像的检测任务需求。

关键词: 模式识别与智能系统, 计算机视觉, 目标检测, 遥感图像, 卷积神经网络

Abstract:

Aiming at the difficulty of object detection caused by the large target size variation， complex background noise and dense targets in airborne wide-area remote sensing images， this paper unifies the target pixel size of the input image by optimizing the segmentation method， and proposes a multi-scale normalized convolutional neural networks model （MNNet）. To enhance the feature correlation between localities， this paper designs a space global connection block （SGC）， which effectively improves the detection accuracy. For the problem that the parameters of the existing NMS algorithm depend on the empirical setting， this paper proposes a self-adaption non-maxima suppression method （DNMS）， which reduces the difficulty of model deployment. The test results on the RSF dataset show that the average precision （AP） of the model in this paper is higher than that of other models by more than 5.0%， and the detection speed reaches 57.7 fps， which can meet the detection task of remote sensing images.

Key words: pattern recognition and intelligent system, computer vision, object detection, remote sensing image, convolutional neural network

中图分类号:

TP391

朱圣杰,王宣,徐芳,彭佳琦,王远超. 机载广域遥感图像的尺度归一化目标检测方法[J]. 吉林大学学报(工学版), 2024, 54(8): 2329-2337.

Sheng-jie ZHU,Xuan WANG,Fang XU,Jia-qi PENG,Yuan-chao WANG. Multi-scale normalized detection method for airborne wide-area remote sensing images[J]. Journal of Jilin University(Engineering and Technology Edition), 2024, 54(8): 2329-2337.

图/表 9

图 1

图 2

图 3

图 4

表 1

表 2

表 3

图 5

图 6

参考文献 24

1	成丽波, 陈鹏宇, 李喆, 等. 基于剪切波变换和拟合优度检验的遥感图像去噪[J]. 吉林大学学报:理学版, 2023, 61(5): 1187-1194.
	Cheng Li-bo, Chen Peng-yu, Li Zhe, JIA Xiaoning. Remote Sensing Image Denoising Based on Shearlet Transform and Goodness of Fit Test[J]. Journal of Jilin University (Science Edition), 2023, 61(5): 1187-1194.
2	成丽波, 董伦, 李喆, 等. 基于NSST与稀疏先验的遥感图像去模糊方法[J]. 吉林大学学报: 理学版, 2024, 62(1): 106-0115.
	Cheng Li-bo, Dong Lun, Li Zhe, et al. Remote Sensing Image Deblurring Method Based on NSST and Sparse Prior[J]. Journal of Jilin University (Science Edition), 2024, 62(1): 106-115.
3	Viola P, Jonus M J. Robust real-time face detection[J]. International Journal of Computer Vision, 2004, 57: 137-154.
4	Dalal N, Triggs B. Histograms of oriented gradients for human detection[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, USA, 2005: 886-893.
5	Girshick R B, Felzenszwalb P F, Mcallester D. Object detection with grammar models[C]//Neural Information Processing Systems, Granada, Spain, 2011: 442-450.
6	Dai J F, Li Y, He K M, et al. R-FCN: object detection via region-based fully convolutional networks[C]//Neural Information Processing Systems (NIPS), Barcelona, Spain, 2016: 379-387.
7	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.
8	高明华, 杨璨. 基于改进卷积神经网络的交通目标检测方法[J]. 吉林大学学报: 工学版, 2022, 52(6): 1353-1361.
	Gao Ming-hua, Yang Can. Traffic target detection method based on improved convolution neural network[J]. Journal of Jilin University (Engineering and Technology Edition), 2022, 52(6): 1353-1361.
9	曲优, 李文辉. 基于锚框变换的单阶段旋转目标检测方法[J]. 吉林大学学报: 工学版, 2022, 52(1): 162-173.
	Qu You, Li Wen-hui. Single-stage rotated object detection network based on anchor transformation[J]. Journal of Jilin University (Engineering and Technology Edition), 2022, 52(1): 162-173.
10	Carion N, Massa F, Synnaeve G, et al. End-to-end object detection with transformers[C]//16th European ECCV Conference, Glasgow, UK, 2020: 213-229.
11	Yang M Y, Liao W T, Li X B, et al. Vehicle detection in aerial images[J]. Photogrammetric Engineering and Remote Sensing, 2019, 85: 297-304.
12	Xia G S, Bai X, Ding J, et al. Dota: a large-scale dataset for object detection in aerial images[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Salt Lake City, USA, 2018: 3974-3983.
13	Van E A. You only look twice: rapid multi-scale object detection in satellite imagery[DB/OL].[2022-12-22]..
14	陈森, 徐伟峰, 王洪涛, 等. 基于改进YOLOv7的麦穗检测算法[J]. 吉林大学学报: 理学版, 2024, 62(4): 886-894.
	Chen Sen, Xu Wei-feng, Wang Hong-tao, et al. Wheat Ear Detection Algorithm Based on Improved YOLOv7[J]. Journal of Jilin University (Science Edition), 2024, 62(4): 886-894.
15	黄键, 徐伟峰, 苏攀, 等. 基于YOLOX-S的车窗状态识别算法[J]. 吉林大学学报: 理学版, 2023, 61(4): 875-882.
	Huang Jian, Xu Wei-feng, Su Pan, et al. Car Window State Recognition Algorithm Based on YOLOX-S[J]. Journal of Jilin University (Science Edition), 2023, 61(4): 875-882.
16	Liu W, Anguelov D, Erhan D, et al. SSD: single shot multibox detector[C]//The 14th European Conference on Computer Vision (ECCV), Amsterdam, Netherlands, 2016: 21-37.
17	He K M, Gkioxari G, Dollár P, et al. Mask R-CNN[C]//IEEE International Conference on Computer Vision (ICCV), Venice, Italy, 2017: 2980-2988.
18	Li K, Wan G, Cheng G, et al. Object detection in optical remote sensing images: a survey and a new benchmark[J]. Photogrammetric Engineering and Remote Sensing, 2020, 159: 296-307.
19	Lu X Q, Zhang Y L, Yuan Y, et al. Gated and axis-concentrated localization network for remote sensing object detection[J]. IEEE Transactions on Remote Sensing, 2020, 58: 179-192.
20	Zou Z X, Shi Z W. 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.
21	Silverman B W. Density estimation for statistics and data analysis[M]. London: Chapman and Hall/CRC, 2018.
22	Zheng Z H, Wang P, Ren D W, et al. Enhancing geometric factors in model learning and inference for object detection and instance segmentation[J]. IEEE Transactions on Cybernetics, 2022, 52(8): 8574-8586.
23	Ma J Q, Shao W Y, Ye H, et al. Arbitrary-oriented scene text detection via rotation proposals[J]. IEEE Transactions on Multimedia, 2018, 20(11): 3111-3122.
24	Yang X, Yang J R, Yan J C, et al. Scrdet: towards more robust detection for small, cluttered and rotated objects[C]//IEEE/CVF International Conference on Computer Vision (ICCV), Seoul, Korea(South), 2019: 8231-8240.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

模型	锚框数	检测头数	参数量	模型大小/MB
YOLOv3	9	3	61 949 149	236.32
YOLOv4	9	3	63 943 071	245.53
YOLOv5m	9	3	22 229 358	84.80
YOLOv5l	9	3	48 384 174	184.57
YOLOv5x	9	3	89 671 790	342.07
SSD300	9	3	23 745 908	90.58
Faster-RCNN	9	3	137 078 239	522.91
MNNet	3	1	31 443 246	119.95

阈值	AP@0.75
阈值	Greedy NMS/%	Soft NMS/%	Softer NMS/%	DNMS/%
0.25	60.65	61.90	61.30	66.37
0.35	62.70	63.50	63.60
0.45	64.10	65.50	65.60
0.55	64.30	65.40	66.00
0.65	64.11	65.55	66.25
0.75	62.95	64.70	64.00
0.85	61.39	62.00	62.70

模型	主干网络	精度/%	召回率/%	F₁值	AP@0.50	AP@0.50∶0.95	帧率/（帧·s^-1）
HOG+SVM	/	6.52	21.19	0.099 7	/	/	1.3 fps
SSD300	VGG-16	25.55	47.34	0.331 8	0.294 6	0.124 5	45.5 fps
R-CenterNet	Hourglass	/	/	/	0.464 0	0.202 1	50.2 fps
RRPN^［23］	VGG-16	48.93	54.72	0.516 6	0.581 3	0.247 2	39.4 fps
SCRDet^［24］	VGG-16	50.37	58.42	0.541 0	0.652 7	0.253 7	27.6 fps
YOLT	Darknet19	22.94	61.23	0.333 8	0.502 2	0.168 1	52.3 fps
R-YOLOv3	Darknet53	21.18	68.65	0.323 7	0.533 4	0.203 4	51.3 fps
R-YOLOv4	CSPDarknet53	39.72	79.25	0.529 1	0.653 1	0.253 8	56.4 fps
R-YOLOv5s	CSPDarknet53	34.28	80.96	0.481 7	0.659 9	0.272 0	71.4 fps
R-YOLOv5m	CSPDarknet53	36.82	79.68	0.503 6	0.635 6	0.283 6	62.1 fps
R-YOLOv5l	CSPDarknet53	40.11	73.21	0.518 2	0.603 3	0.235 4	48.5 fps
MNNet（without SGC）	CSPDarknet53	51.34	74.12	0.606 6	0.653 6	0.293 4	68.9 fps
MNNet	CSPDarknet53	59.79	71.17	0.649 8	0.717 9	0.341 2	57.7 fps

[1]	才华,寇婷婷,杨依宁,马智勇,王伟刚,孙俊喜. 基于轨迹优化的三维车辆多目标跟踪[J]. 吉林大学学报(工学版), 2024, 54(8): 2338-2347.
[2]	特木尔朝鲁朝鲁,张亚萍. 基于卷积神经网络的无线传感器网络链路异常检测算法[J]. 吉林大学学报(工学版), 2024, 54(8): 2295-2300.
[3]	赵宏伟,武鸿,马克,李海. 基于知识蒸馏的图像分类框架[J]. 吉林大学学报(工学版), 2024, 54(8): 2307-2312.
[4]	张锦洲,姬世青,谭创. 融合卷积神经网络和双边滤波的相贯线焊缝提取算法[J]. 吉林大学学报(工学版), 2024, 54(8): 2313-2318.
[5]	游新冬,郭磊,韩晶,吕学强. 一种工件表面压印字符识别网络[J]. 吉林大学学报(工学版), 2024, 54(7): 2072-2079.
[6]	孙铭会,薛浩,金玉波,曲卫东,秦贵和. 联合时空注意力的视频显著性预测[J]. 吉林大学学报(工学版), 2024, 54(6): 1767-1776.
[7]	魏晓辉,王晨洋,吴旗,郑新阳,于洪梅,岳恒山. 面向脉动阵列神经网络加速器的软错误近似容错设计[J]. 吉林大学学报(工学版), 2024, 54(6): 1746-1755.
[8]	王殿伟,张池,房杰,许志杰. 基于高分辨率孪生网络的无人机目标跟踪算法[J]. 吉林大学学报(工学版), 2024, 54(5): 1426-1434.
[9]	王宇,赵凯. 基于亚像素定位的人体姿态热图后处理[J]. 吉林大学学报(工学版), 2024, 54(5): 1385-1392.
[10]	高云龙,任明,吴川,高文. 基于注意力机制改进的无锚框舰船检测模型[J]. 吉林大学学报(工学版), 2024, 54(5): 1407-1416.
[11]	夏超,王梦佳,朱剑月,杨志刚. 基于分层卷积自编码器的钝体湍流流场降阶分析[J]. 吉林大学学报(工学版), 2024, 54(4): 874-882.
[12]	陈仁祥,胡超超,胡小林,杨黎霞,张军,何家乐. 基于改进YOLOv5的驾驶员分心驾驶检测[J]. 吉林大学学报(工学版), 2024, 54(4): 959-968.
[13]	张云佐,郭威,李文博. 遥感图像密集小目标全方位精准检测算法[J]. 吉林大学学报(工学版), 2024, 54(4): 1105-1113.
[14]	王宏志,宋明轩,程超,解东旋. 基于改进YOLOv4-tiny算法的车距预警方法[J]. 吉林大学学报(工学版), 2024, 54(3): 741-748.
[15]	李雄飞,宋紫萱,朱芮,张小利. 基于多尺度融合的遥感图像变化检测模型[J]. 吉林大学学报(工学版), 2024, 54(2): 516-523.

机载广域遥感图像的尺度归一化目标检测方法

Multi-scale normalized detection method for airborne wide-area remote sensing images

RICH HTML

PDF (PC)

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 24

相关文章 15

Metrics

本文评价

推荐阅读 10