一種目標(biāo)區(qū)域特征增強(qiáng)的SAR圖像飛機(jī)目標(biāo)檢測與識(shí)別網(wǎng)絡(luò)

韓萍; 趙涵; 廖大鈺; 彭彥文; 程爭

doi:10.11999/JEIT240491

一種目標(biāo)區(qū)域特征增強(qiáng)的SAR圖像飛機(jī)目標(biāo)檢測與識(shí)別網(wǎng)絡(luò)

doi: 10.11999/JEIT240491 cstr: 32379.14.JEIT240491

中國民航大學(xué)智能信號(hào)與圖像處理天津市重點(diǎn)實(shí)驗(yàn)室天津 300300

基金項(xiàng)目: 中央高?；?3122020043)

詳細(xì)信息

作者簡介:
韓萍：女，教授，研究方向?yàn)镾AR圖像處理與目標(biāo)檢測

趙涵：男，碩士生，研究方向?yàn)镾AR圖像飛目標(biāo)檢測

廖大鈺：男，碩士生，研究方向?yàn)镾AR圖像飛機(jī)目標(biāo)檢測

彭彥文：男，碩士生，研究方向?yàn)镻oLSAR圖像飛機(jī)場跑道檢測

程爭：男，實(shí)驗(yàn)師，研究方向?yàn)闃O化SAR圖像處理與目標(biāo)檢測

通訊作者:
韓萍　hanpingcauc@163.com

中圖分類號(hào): TN958
計(jì)量
- 文章訪問數(shù): 545
- HTML全文瀏覽量: 239
- PDF下載量: 92
- 被引次數(shù): 0
出版歷程
- 收稿日期: 2024-06-14
- 修回日期: 2024-11-21
- 網(wǎng)絡(luò)出版日期: 2024-11-25
- 刊出日期: 2024-12-01

A SAR Image Aircraft Target Detection and Recognition Network with Target Region Feature Enhancement

Tianjin Key Laboratory of Intelligent Signal and Image Processing, Civil Aviation University of China, Tianjin 300300, China

Funds: The Central University Fund (3122020043)

摘要

摘要: 在合成孔徑雷達(dá)(SAR)圖像飛機(jī)目標(biāo)檢測識(shí)別中，飛機(jī)目標(biāo)圖像呈現(xiàn)離散特性以及結(jié)構(gòu)之間的相似性會(huì)降低飛機(jī)檢測與識(shí)別的準(zhǔn)確率。為此該文設(shè)計(jì)了一種目標(biāo)區(qū)域特征增強(qiáng)的SAR圖像飛機(jī)目標(biāo)檢測與識(shí)別網(wǎng)絡(luò)。網(wǎng)絡(luò)由3部分組成：保護(hù)飛機(jī)特征的跨階段部分網(wǎng)絡(luò)(FP-CSPDarnet)、自適應(yīng)特征融合的特征金字塔(FPN-A)以及目標(biāo)區(qū)域散射特征提取與增強(qiáng)的檢測頭(D-Head)。FP-CSPDarnet在提取特征的同時(shí)可以有效保護(hù)SAR圖像飛機(jī)特征；FPN-A采用多層次特征自適應(yīng)融合、細(xì)化，來增強(qiáng)飛機(jī)特征；D-Head在檢測前有效增強(qiáng)飛機(jī)可辨別特征，提升飛機(jī)檢測與識(shí)別精度。利用SAR-ADRD數(shù)據(jù)集的實(shí)驗(yàn)結(jié)果證明了該文所提方法有效性，其平均精度相對與基線網(wǎng)絡(luò)YOLOv5s提升了2.0%。
- 合成孔徑雷達(dá) /
- 飛機(jī)目標(biāo)檢測與識(shí)別 /
- YOLOv5s /
- 飛機(jī)特征保護(hù) /
- 特征增強(qiáng)
Abstract: In Synthetic Aperture Radar (SAR) image aircraft target detection and recognition, the discrete characteristics of aircraft target images and the similarity between structures can reduce the accuracy of aircraft detection and recognition. A SAR image aircraft target detection and recognition network with enhanced target area features is proposed in this paper. The network consists of three parts: Feature Protecting Cross Stage Partial Darknet (FP-CSPDarnet) for protecting aircraft features, Feature Pyramid Net with Adaptive fusion (FPN-A) for adaptive feature fusion, and Detection Head for target area scattering feature extraction and enhancement (D-Head). FP-CSPDarnet can effectively protect the aircraft features in SAR images while extracting features; FPN-A adopts multi-level feature adaptive fusion and refinement to enhance aircraft features; D-Head effectively enhances the identifiable features of the aircraft before detection, improving the accuracy of aircraft detection and recognition. The experimental results using the SAR-ADRD dataset have demonstrated the effectiveness of the proposed method, with an average accuracy improvement of 2.0% compared to the baseline network YOLOv5s.
- Synthetic Aperture Radar(SAR) /
- Aircraft target detection and recognition /
- YOLOv5s /
- Aircraft feature protection /
- Target area feature enhancement

HTML全文

圖 1 SAR圖像中飛機(jī)的結(jié)構(gòu)展示

下載: 全尺寸圖片幻燈片

圖 2 YOLOv5網(wǎng)絡(luò)架構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 3 FPN架構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 4 本文網(wǎng)絡(luò)架構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 5 CSPDarknet骨干網(wǎng)絡(luò)與FP-CSPDarnet結(jié)構(gòu)對比圖

下載: 全尺寸圖片幻燈片

圖 6 SPD-Conv結(jié)構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 7 FPN-A結(jié)構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 8 AR模塊

下載: 全尺寸圖片幻燈片

圖 9 D-Head結(jié)構(gòu)圖

下載: 全尺寸圖片幻燈片

圖 10 AF模塊

下載: 全尺寸圖片幻燈片

圖 11 數(shù)據(jù)集中飛機(jī)的類別與數(shù)量

下載: 全尺寸圖片幻燈片

圖 12 骨干網(wǎng)絡(luò)輸出特征圖通道可視化結(jié)果圖

下載: 全尺寸圖片幻燈片

圖 13 優(yōu)化檢測頭前后的混淆矩陣

下載: 全尺寸圖片幻燈片

圖 14 各個(gè)網(wǎng)絡(luò)檢測結(jié)果圖1

下載: 全尺寸圖片幻燈片

圖 15 各個(gè)網(wǎng)絡(luò)檢測結(jié)果圖2

下載: 全尺寸圖片幻燈片

表 1 YOLOv5網(wǎng)絡(luò)深度消融實(shí)驗(yàn)

模型	P(%)	R(%)	mAP(%)	Parameters(M)	fps
YOLOv5n	90.6	89.1	92.0	1.8	93.6
YOLOv5s	93.0	90.6	93.7	7.0	93.5
YOLOv5m	92.2	90.2	93.3	20.9	68.5
YOLOv5l	93.1	91.2	93.0	46.1	47.6
YOLOv5x	94.0	91.0	94.4	86.2	41.9

下載: 導(dǎo)出CSV

表 2 網(wǎng)絡(luò)各個(gè)模塊消融實(shí)驗(yàn)

	FPN-A	FP-CSPDarnet	D-Head	P(%)	R(%)	mAP(%)	Parameters(M)	fps
Baseline	–	–	–	93.0	90.6	93.7	7.0	93.6
	√	–	–	90.2	86.5	92.4	8.5	129.9
	√	√	–	92.3	92.5	94.8	10.3	129.9
本文方法	√	√	√	92.5	92.3	95.7	12.2	108.7

下載: 導(dǎo)出CSV

表 3 骨干網(wǎng)絡(luò)P-CSPDarknet中各模塊消融實(shí)驗(yàn)(%)

FPN-A	FP-CSPDarnet (骨干結(jié)構(gòu))	FP-CSPDarnet (骨干結(jié)構(gòu)+SPD-Conv)	P (%)	R (%)	mAP
√	–	–	90.2	86.5	92.4
√	√	–	91.8	90.2	93.9
√	√	√	92.3	92.5	94.8

下載: 導(dǎo)出CSV

表 4 不同檢測網(wǎng)絡(luò)對比實(shí)驗(yàn)

	P(%)	R(%)	mAP(%)	Parameters(M)	fps
YOLOv5s	92.6	89.9	93.7	7.0	93.5
Faster R-CNN	82.0	85.6	87.8	41.2	11.2
TOOD	84.9	81.7	85.0	31.8	12.9
YOLOX-s	80.7	83.4	89.7	8.9	41.2
YOLOv7s	91.1	87.6	93.5	9.2	75.8
本文方法	92.5	92.3	95.7	12.3	108.7

下載: 導(dǎo)出CSV

表 5 數(shù)據(jù)集內(nèi)不同飛機(jī)類別在不同檢測網(wǎng)絡(luò)的精度(%)

網(wǎng)絡(luò)模型/飛機(jī)類別	Boeing787	A220	ARJ21	Boeing737-800	A320/321	A330	Others	mAP(%)
YOLOv5s	98.0	96.6	93.6	87.2	88.1	95.1	96.9	93.7
Yolov7s	96.8	95.5	92.2	93.0	85.8	96.2	94.2	93.5
TOOD	90.7	94.3	85.3	81.1	67.8	91.0	84.4	84.9
YOLOX-s	89.2	86.4	86.3	94.5	95.4	86.7	89.5	89.7
Faster R-CNN	91.8	94.8	87.9	85.6	74.7	91.1	89.0	87.8
本文方法	98.7	98.7	97.7	95.1	95.4	94.8	97.1	95.7

下載: 導(dǎo)出CSV

參考文獻(xiàn)(24)

[1]	高貴, 周蝶飛, 蔣詠梅, 等. SAR圖像目標(biāo)檢測研究綜述[J]. 信號(hào)處理, 2008, 24(6): 971–981. doi: 10.3969/j.issn.1003-0530.2008.06.018. GAO Gui, ZHOU Diefei, JIANG Yongmei, et al. Study on target detection in SAR Image: A survey[J]. Signal Processing, 2008, 24(6): 971–981. doi: 10.3969/j.issn.1003-0530.2008.06.018.
[2]	李永禎, 黃大通, 邢世其, 等. 合成孔徑雷達(dá)干擾技術(shù)研究綜述[J]. 雷達(dá)學(xué)報(bào), 2020, 9(5): 753–764. doi: 10.12000/JR20087. LI Yongzhen, HUANG Datong, XING Shiqi, et al. A review of synthetic aperture radar jamming technique[J]. Journal of Radars, 2020, 9(5): 753–764. doi: 10.12000/JR20087.
[3]	FU Kun, DOU Fangzheng, LI Hengchao, et al. Aircraft recognition in SAR images based on scattering structure feature and template matching[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2018, 11(11): 4206–4217. doi: 10.1109/JSTARS.2018.2872018.
[4]	HU Hao, HUANG Lanqing, and YU Wenxian. Aircraft detection for HR SAR Images in non-homogeneous background using GGMD-based modeling[J]. Chinese Journal of Electronics, 2019, 28(6): 1271–1280. doi: 10.1049/cje.2019.08.010.
[5]	CHEN Jiehong, ZHANG Bo, and WANG Chao. Backscattering feature analysis and recognition of civilian aircraft in TerraSAR-X images[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12(4): 796–800. doi: 10.1109/LGRS.2014.2362845.
[6]	HE Chu, TU Mingxia, LIU Xinlong, et al. Mixture statistical distribution based multiple component model for target detection in high resolution SAR imagery[J]. ISPRS International Journal of Geo-Information, 2017, 6(11): 336. doi: 10.3390/ijgi6110336.
[7]	高君, 高鑫, 孫顯. 基于幾何特征的高分辨率SAR圖像飛機(jī)目標(biāo)解譯方法[J]. 國外電子測量技術(shù), 2015, 34(8): 21–28. doi: 10.3969/j.issn.1002-8978.2015.08.008. GAO Jun, GAO Xin, and SUN Xian. Geometrical features-based method for aircraft target interpretation in high-resolution SAR images[J]. Foreign Electronic Measurement Technology, 2015, 34(8): 21–28. doi: 10.3969/j.issn.1002-8978.2015.08.008.
[8]	ZHANG Peng, XU Hao, TIAN Tian, et al. SFRE-net: Scattering feature relation enhancement network for aircraft detection in SAR images[J]. Remote Sensing, 2022, 14(9): 2076. doi: 10.3390/rs14092076.
[9]	趙琰. 基于深度學(xué)習(xí)的SAR圖像飛機(jī)目標(biāo)檢測與識(shí)別[D]. [碩士論文], 國防科技大學(xué), 2020. doi: 10.27052/d.cnki.gzjgu.2020.001038. ZHAO Yan. Deep learning based aircraft detection and recognition in SAR images[D]. [Master dissertation], National University of Defense Technology, 2020. doi: 10.27052/d.cnki.gzjgu.2020.001038.
[10]	WANG Zhen, XU Nan, GUO Jianxin, et al. SCFNet: Semantic condition constraint guided feature aware network for aircraft detection in SAR Images[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5239420. doi: 10.1109/TGRS.2022.3224599.
[11]	王智睿, 康玉卓, 曾璇, 等. SAR-AIRcraft-1.0: 高分辨率SAR飛機(jī)檢測識(shí)別數(shù)據(jù)集[J]. 雷達(dá)學(xué)報(bào), 2023, 12(4): 906–922. doi: 10.12000/JR23043. WANG Zhirui, KANG Yuzhuo, ZENG Xuan, et al. SAR-AIRcraft-1.0: High-resolution SAR aircraft detection and recognition dataset[J]. Journal of Radars, 2023, 12(4): 906–922. doi: 10.12000/JR23043.
[12]	WU Wentong, LIU Han, LI Lingling, et al. Application of local fully convolutional neural network combined with YOLO v5 algorithm in small target detection of remote sensing image[J]. PLoS One, 2021, 16(10): e0259283. doi: 10.1371/journal.pone.0259283.
[13]	LIN T Y, DOLLáR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Hawaii, American, 2017: 2117–2125. doi: 10.1109/CVPR.2017.106.
[14]	CHEN Qiang, WANG Yingming, YANG Tong, et al. You only look one-level feature[C]. 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Nashville, USA, 2021: 13039–13048. doi: 10.1109/CVPR46437.2021.01284.
[15]	SONG Guanglu, LIU Yu, and WANG Xiaogang. Revisiting the sibling head in object detector[C]. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, American, 2020: 11563–11572. doi: 10.1109/CVPR42600.2020.01158.
[16]	ZHUANG Jiayuan, QIN Zheng, YU Hao, et al. Task-specific context decoupling for object detection[J]. arXiv preprint arXiv: 2303.01047, 2023.
[17]	GE Zhang, LIU Songtao, WANG Feng, et al. YOLOX: Exceeding YOLO series in 2021[J]. arXiv preprint arXiv: 2107.08430, 2021.
[18]	FENG Chengjian, ZHONG Yujie, GAO Yu, et al. TOOD: Task-aligned one-stage object detection[C]. 2021 IEEE/CVF International Conference on Computer Vision, Montreal, Canada, 2021: 3490–3499. doi: 10.1109/ICCV48922.2021.00349.
[19]	LI Chuyi, LI Lulu, JIANG Hongliang, et al. YOLOv6: A single-stage object detection framework for industrial applications[J]. arXiv preprint arXiv: 2209.02976, 2022.
[20]	WANG Gang, CHEN Yanfei, AN Pei, et al. UAV-YOLOv8: A small-object-detection model based on improved YOLOv8 for UAV aerial photography scenarios[J]. Sensors, 2023, 23(16): 7190. doi: 10.3390/s23167190.
[21]	ZHAO Yan, ZHAO Lingjun, LIU Zhong, et al. Attentional feature refinement and alignment network for aircraft detection in SAR imagery[J]. IEEE Transactions on Geoscience and Remote Sensing, 2022, 60: 5220616. doi: 10.1109/TGRS.2021.3139994.
[22]	HAN Ping, LIAO Dayu, HAN Binbin, et al. SEAN: A simple and efficient attention network for aircraft detection in SAR images[J]. Remote Sensing, 2022, 14(18): 4669. doi: 10.3390/rs14184669.
[23]	REN Shaoqing, HE Kaiming, 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.
[24]	WANG C Y, BOCHKOVSKIY A, and LIAO H Y M. YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors[C]. 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition, Vancouver, Canada, 2023: 7464–7475. doi: 10.1109/CVPR52729.2023.00721.