基于全卷积神经网络和低分辨率标签的森林变化检测研究

doi:10.12302/j.issn.1000-2006.202204069

国家林草科技领军期刊
中国精品科技期刊
中国高校百佳科技期刊
江苏省新闻出版政府奖期刊奖
RCCSE林学权威期刊（A+）
CSCD核心期刊
Scopus数据库收录期刊
中文核心期刊
SCD核心期刊

作者加群：102861116

微信公众号：南京林业大学学报

高级检索

南京林业大学学报（自然科学版） ›› 2024, Vol. 48 ›› Issue (1): 187-195.doi: 10.12302/j.issn.1000-2006.202204069

基于全卷积神经网络和低分辨率标签的森林变化检测研究

向俊¹^,²(), 严恩萍², 姜镓伟²^,³, 宋亚斌⁴, 韦维¹^,^*(), 莫登奎²^,^*()

1.广西壮族自治区林业科学研究院,广西南宁 530002
2.林业遥感大数据与生态安全湖南省重点实验室,中南林业科技大学林学院,湖南长沙 410004
3.中山大学土木工程学院,广东珠海 519082
4.国家林业和草原局中南调查规划设计院,湖南长沙 410014

收稿日期:2022-04-29 修回日期:2022-06-19 出版日期:2024-01-30 发布日期:2024-01-24
通讯作者: 韦维,莫登奎
基金资助:
国家自然科学基金项目(32071682);国家自然科学基金项目(31901311);国家林业和草原局中南调查规划设计院项目(68218022);湖南省林业科技创新计划(XLK202108-8)

Research on forest change detection based on fully convolutional network and low resolution label

XIANG Jun¹^,²(), YAN Enping², JIANG Jiawei²^,³, SONG Yabin⁴, WEI Wei¹^,^*(), MO Dengkui²^,^*()

1. Guangxi Zhuang Autonomous Region Forestry Research Institute, Nanning 530002, China
2. Hunan Provincial Key Laboratory of Forestry Remote Sensing Big Data and Ecological Security, College of Forestry, Central South University of Forestry and Technology, Changsha 410004, China
3. School of Civil Engineering, Sun Yat-Sen University, Zhuhai 519082, China
4. Central South Inventory and Planning Institute of National Forestry and Grassland Administration, Changsha 410014, China

Received:2022-04-29 Revised:2022-06-19 Online:2024-01-30 Published:2024-01-24
Contact: WEI Wei,MO Dengkui

摘要/Abstract

摘要：

【目的】针对目前森林变化检测中高精度标签样本缺失或不足的问题,提出一种基于全卷积神经网络和低分辨率标签的森林变化检测方法,旨在实现林地区域内森林变化的简易快速提取。【方法】首先对获取的数据进行去云、筛选、标签融合等预处理,利用全卷积神经网络模型分别提取2020年和2021年研究区森林高分遥感影像,并评价模型精度;利用分类后比较法获取森林变化区域,得到变化结果并与目视解译结果进行对比,基于像素面积计算森林变化检测的精确率等评价指标。【结果】所用全卷积神经网络(FCN)模型在2020年森林提取结果的精确率和召回率的调和均值(F₁分数)为97.09%,2021年森林提取结果的F₁分数为95.96%,与分割网络模型(U-Net、FPN、LinkNet)相比更优。比较两期森林提取结果得到变化区域,森林增加与森林减少的合计变化精确率为73.30%,召回率为77.37%,F₁分数为75.28%。【结论】该方法实现了基于低分辨率标签对高分遥感影像森林变化区域进行快速、准确的获取。采用少量的低分辨率标签完成森林变化检测任务,同时可为大面积林地变更调查提供参考。

关键词: 低分辨率标签, 全卷积神经网络, 深度学习, 森林变化检测

Abstract:

【Objective】A forest change detection method based on fully convolutional networks and low resolution labels is proposed to address the problem of missing or insufficient high-precision label samples in current forest change detection, with the goal of achieving simple and rapid extraction of forest changes in forest areas. 【Method】First, the gathered data was de-clouded, screened and labeled, and then the fully convolutional network model was used to extract the forests in high-scoring remote sensing photos in the study area in 2020 and 2021, respectively, and the model accuracy was evaluated. The forest change area was calculated using the post-classification comparison method, and the findings were compared with visual interpretation results. The pixel area was used to calculate evaluation indicators, such as forest change detection accuracy. 【Result】Experiments reveal that the F₁ score of the model employed in this research is 97.09% in 2020 forest extraction results and 95.96% in 2021 forest extraction results, which was the best among segmentation network models (U-Net, FPN, LinkNet). The total change precision rate of forest increase and forest decline was 73.30%, the recall rate was 77.37%, and the F₁ score was 75.28% when comparing the forest extraction data from the two periods to obtain the changed area. 【Conclusion】Based on low resolution labeling, this method allows for the speedy and precise capture of forest change regions from high-resolution remote sensing pictures. To accomplish forest change detection, a small number of low-resolution labels are used, which can also serve as a reference for large-scale forestland change inquiries.

Key words: low resolution label, fully convolutional network, deep learning, forest changes detection

中图分类号:

TP391.4
S757

向俊,严恩萍,姜镓伟,等. 基于全卷积神经网络和低分辨率标签的森林变化检测研究[J]. 南京林业大学学报（自然科学版）, 2024, 48(1): 187-195.

XIANG Jun, YAN Enping, JIANG Jiawei, SONG Yabin, WEI Wei, MO Dengkui. Research on forest change detection based on fully convolutional network and low resolution label[J].Journal of Nanjing Forestry University (Natural Science Edition), 2024, 48(1): 187-195.DOI: 10.12302/j.issn.1000-2006.202204069.

图/表 9

图1

图2

图3

图4

图5

表1

图6

图7

表2

参考文献 37

[1]	魏安世, 杨志刚. 森林资源年度监测小班数据自动更新技术[J]. 南京林业大学学报(自然科学版), 2010, 34(4):123-128.
	WEI A S, YANG Z G. Automatic updating technique of subcompartment data for annual monitoring of forest resource[J]. J Nanjing For Univ (Nat Sci Ed), 2010, 34(4):123-128.DOI: 10.3969/j.issn.1000-2006.2010.04.027.
[2]	刘羿, 佘光辉, 刘安兴, 等. 森林资源系统自组织特征研究[J]. 南京林业大学学报(自然科学版), 2008, 32(5):51-55.
	LIU Y, SHE G H, LIU A X, et al. Research on self-organization characters in forest resource system[J]. J Nanjing For Univ (Nat Sci Ed), 2008, 32(5):51-55.DOI: 10.3969/j.issn.1000-2006.2008.05.012.
[3]	李春干, 梁文海. 基于面向对象变化向量分析法的遥感影像森林变化检测[J]. 国土资源遥感, 2017, 29(3):77-84.
	LI C G, LIANG W H. Forest change detection using remote sensing image based on object-oriented change vector analysis[J]. Remote Sens Land Resour, 2017, 29(3):77-84.DOI: 10.6046/gtzyyg.2017.03.11.
[4]	张丽云, 赵天忠, 夏朝宗, 等. 遥感变化检测技术在林业中的应用[J]. 世界林业研究, 2016, 29(2):44-48.
	ZHANG L Y, ZHAO T Z, XIA C Z, et al. Application of change detection technologies of remote sensing to forestry[J]. World For Res, 2016, 29(2):44-48.DOI: 10.13348/j.cnki.sjlyyj.2016.02.005.
[5]	张祖宇, 滕永核, 秦元丽, 等. 基于U-Net模型的无人机影像数据地表覆被信息自动提取研究[J]. 广西林业科学, 2022, 51(4):516-519.
	ZHANG Z Y, TENG Y H, QIN Y L, et al. Automatic extraction of land cover information from UAV image data based on U-Net model[J]. Guangxi Forest Sci, 2022, 51(4):516-519.DOI: 10.19692/j.issn.1006-1126.20220411.
[6]	王利民, 刘佳, 杨玲波, 等. 基于无人机影像的农情遥感监测应用[J]. 农业工程学报, 2013, 29(18):136-145.
	WANG L M, LIU J, YANG L B, et al. Applications of unmanned aerial vehicle images on agricultural remote sensing monitoring[J]. Trans Chin Soc Agric Eng, 2013, 29(18):136-145.DOI: 10.3969/j.issn.1002-6819.2013.18.017.
[7]	ZHAO S H, WANG Q, LI Y, et al. An overview of satellite remote sensing technology used in China’s environmental protection[J]. Earth Sci Inform, 2017, 10(2):137-148.DOI: 10.1007/s12145-017-0286-6.
[8]	ROGAN J, CHEN D M. Remote sensing technology for mapping and monitoring land-cover and land-use change[J]. Prog Plan, 2004, 61(4):301-325.DOI: 10.1016/S0305-9006(03)00066-7.
[9]	郭颖, 李增元, 陈尔学, 等. 一种改进的高空间分辨率遥感影像森林类型深度学习精细分类方法:双支FCN-8s[J]. 林业科学, 2020, 56(3):48-60.
	GUO Y, LI Z Y, CHEN E X, et al. A deep learning method for forest fine classification based on high resolution remote sensing images:two-branch FCN-8s[J]. Sci Silvae Sin, 2020, 56(3):48-60.DOI: 10.11707/j.1001-7488.20200306.
[10]	覃先林, 李晓彤, 刘树超, 等. 中国林火卫星遥感预警监测技术研究进展[J]. 遥感学报, 2020, 24(5):511-520.
	QIN X L, LI X T, LIU S C, et al. Forest fire early warning and monitoring techniques using satellite remote sensing in China[J]. J Remote Sens, 2020, 24(5):511-520.
[11]	杨雷, 禹定峰, 高皜, 等. Sentinel-2的胶州湾水体透明度遥感反演[J]. 红外与激光工程, 2021, 50(12):515-521.
	YANG L, YU D F, GAO H, et al. Remote sensing retrieval of secchi disk depth in Jiaozhou Bay using Sentinel-2 MSI image[J]. Infrared Laser Eng, 2021, 50(12):515-521.
[12]	陈锐志, 王磊, 李德仁, 等. 导航与遥感技术融合综述[J]. 测绘学报, 2019, 48(12):1507-1522.
	CHEN R Z, WANG L, LI D R, et al. A survey on the fusion of the navigation and the remote sensing techniques[J]. Acta Geod Cartogr Sin, 2019, 48(12):1507-1522.DOI: 10.11947/j.AGCS.2019.20190446.
[13]	GU J X, WANG Z H, KUEN J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognit, 2018, 77:354-377.DOI: 10.1016/j.patcog.2017.10.013.
[14]	SHELHAMER E, LONG J, DARRELL T. Fully convolutional networks for semantic segmentation[C]// IEEE Transactions on Pattern Analysis and Machine Intelligence.May 24,2016,IEEE, 2016:640-651.DOI: 10.1109/TPAMI.2016.2572683.
[15]	业巧林, 许平, 张冬. 基于深度学习特征和支持向量机的遥感图像分类[J]. 林业工程学报, 2019, 4(2):119-125.
	YE Q L, XU D P, ZHANG D. Remote sensing image classification based on deep learning features and support vector machine[J]. J Fore Eng, 2019, 4(2):119-125.DOI:10.13360/j.issn.2096-1359.2019.02.019.
[16]	FU G, LIU C J, ZHOU R, et al. Classification for high resolution remote sensing imagery using a fully convolutional network[J]. Remote Sens, 2017, 9(5):498.DOI: 10.3390/rs9050498.
[17]	LIU R C, JIANG D W, ZHANG L L, et al. Deep depthwise separable convolutional network for change detection in optical aerial images[J]. IEEE J Sel Top Appl Earth Obs Remote Sens, 2020, 13:1109-1118.DOI: 10.1109/JSTARS.2020.2974276.
[18]	LEI T, ZHANG Q, XUE D H, et al. End-to-end change detection using a symmetric fully convolutional network for landslide mapping[C]// ICASSP 2019—2019 IEEE International Conference on Acoustics,Speech and Signal Processing (ICASSP). May 12-17,2019,Brighton,UK.IEEE, 2019:3027-3031.DOI: 10.1109/ICASSP.2019.8682802.
[19]	MILLETARI F, NAVAB N, AHMADI S A. V-net:fully convolutional neural networks for volumetric medical image segmentation[C]// 2016 Fourth International Conference on 3D Vision (3DV). October 25-28,2016,Stanford,CA,USA.IEEE, 2016:565-571.DOI: 10.1109/3DV.2016.79.
[20]	RONNEBERGER O, FISCHER P, BROX T. U-net:convolutional networks for biomedical image segmentation[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention. Cham:Springer, 2015:234-241.DOI:10.1007/978-3-319-24574-4_28.
[21]	CHEN H, SHI Z W. A spatial-temporal attention-based method and a new dataset for remote sensing image change detection[J]. Remote Sens, 2020, 12(10):1662.DOI: 10.3390/rs12101662.
[22]	王昶, 张永生, 王旭, 等. 基于深度学习的遥感影像变化检测方法[J]. 浙江大学学报(工学版), 2020, 54(11):2138-2148.
	WANG C, ZHANG Y S, WANG X, et al. Remote sensing image change detection method based on deep neural networks[J]. J Zhejiang Univ (Eng Sci), 2020, 54(11):2138-2148.DOI: 10.3785/j.issn.1008-973X.2020.11.009.
[23]	WANG Z Y, LIU M L, LIU X N, et al. Spatio-temporal evolution of surface urban heat islands in the Chang-Zhu-Tan urban agglomeration[J]. Phys Chem Earth Parts A/B/C, 2020, 117:102865.DOI: 10.1016/j.pce.2020.102865.
[24]	JIANG W G, DENG Y, TANG Z H, et al. Modelling the potential impacts of urban ecosystem changes on carbon storage under different scenarios by linking the CLUE-S and the InVEST models[J]. Ecol Model, 2017, 345:30-40.DOI: 10.1016/j.ecolmodel.2016.12.002.
[25]	马润, 胡斯勒图, 尚华哲, 等. 基于葵花-8卫星大气产品的地表下行短波辐射计算[J]. 遥感学报, 2019, 23(5):924-934.
	MA R, HUSI L, SHANG H Z, et al. Estimation of downward surface shortwave radiation from Himawari-8 atmospheric products[J]. J Remote Sens, 2019, 23(5):924-934.
[26]	伟乐斯, 尚华哲, 胡斯勒图, 等. GF-5 DPC数据的云检测方法研究[J]. 遥感学报, 2021, 25(10):2053-2066.
	WEI L S, SHANG H Z, HUSI L, et al. Cloud detection algorithm based on GF-5 DPC data[J]. J Remote Sens, 2021, 25(10):2053-2066.
[27]	PONTIUS R G, SHUSAS E, MCEACHERN M. Detecting important categorical land changes while accounting for persistence[J]. Agric Ecosyst Environ, 2004, 101(2/3):251-268.DOI: 10.1016/j.agee.2003.09.008.
[28]	KINGMA D P, BA J. Adam:a method for stochastic optimization[EB/OL]. arXiv:Learning, 2014.[2022-03-20]. https://arxiv.org/abs/1412.6980.
[29]	HAND D, CHRISTEN P. A note on using the F-measure for evaluating record linkage algorithms[J]. Stat Comput, 2018, 28(3):539-547.DOI: 10.1007/s11222-017-9746-6.
[30]	YIN H, PFLUGMACHER D, LI A, et al. Land use and land cover change in Inner Mongolia-understanding the effects of China’s re-vegetation programs[J]. Remote Sens Environ, 2018, 204:918-930.DOI: 10.1016/j.rse.2017.08.030.
[31]	JIN S M, YANG L M, ZHU Z, et al. A land cover change detection and classification protocol for updating Alaska NLCD 2001 to 2011[J]. Remote Sens Environ, 2017, 195:44-55.DOI: 10.1016/j.rse.2017.04.021.
[32]	LI J Y, HUANG X, CHANG X Y. A label-noise robust active learning sample collection method for multi-temporal urban land-cover classification and change analysis[J]. ISPRS J Photogramm Remote Sens, 2020, 163:1-17.DOI: 10.1016/j.isprsjprs.2020.02.022.
[33]	HARALICK R M, STERNBERG S R, ZHUANG X H. Image analysis using mathematical morphology[J]. IEEE Trans Pattern Anal Mach Intell, 1987, PAMI-9(4):532-550.DOI: 10.1109/TPAMI.1987.4767941.
[34]	吴胜义, 张方圆, 王飞. 林地变更调查技术方法分析与研究[J]. 林业科技, 2021, 46(2):38-41,45.
	WU S Y, ZHANG F Y, WANG F. Analysis and research on technology and method of forest land change survey[J]. For Sci Technol, 2021, 46(2):38-41,45.DOI: 10.19750/j.cnki.1001-9499.2021.02.011.
[35]	徐新良, 刘纪远, 庄大方, 等. 中国林地资源时空动态特征及驱动力分析[J]. 北京林业大学学报, 2004, 26(1):41-46.
	XU X L, LIU J Y, ZHUANG D F, et al. Spatial-temporal characteristics and driving forces of woodland resource changes in China[J]. J Beijing For Univ, 2004, 26(1):41-46.DOI: 10.3321/j.issn:1000-1522.2004.01.008.
[36]	夏传福, 李静, 柳钦火. 植被物候遥感监测研究进展[J]. 遥感学报, 2013, 17(1):1-16.
	XIA C F, LI J, LIU Q H. Review of advances in vegetation phenology monitoring by remote sensing[J]. J Remote Sens, 2013, 17(1):1-16.
[37]	范德芹, 赵学胜, 朱文泉, 等. 植物物候遥感监测精度影响因素研究综述[J]. 地理科学进展, 2016, 35(3):304-319.
	FAN D Q, ZHAO X S, ZHU W Q, et al. Review of influencing factors of accuracy of plant phenology monitoring based on remote sensing data[J]. Prog Geogr, 2016, 35(3):304-319.

森林检测模型 forest detection model	年份 year	样本总面积/ km² total sample area	森林面积/ km² forest area	S_TP/km²	S_FP/km²	S_FN/km²	精确率/% precision	召回率/% recall	F₁分数/% F₁ score
FPN	2020	400	347.07	338.25	21.77	8.82	93.95	97.46	95.67
	2021	400	344.78	335.57	24.02	9.21	93.32	97.33	95.28
LinkNet	2020	400	347.07	336.87	22.65	10.20	93.70	97.06	95.35
	2021	400	344.78	328.27	14.90	16.51	95.66	95.21	95.43
U-Net	2020	400	347.07	339.36	18.55	7.71	94.82	97.78	96.28
	2021	400	344.78	314.65	11.36	30.13	96.52	91.26	93.81
FCN	2020	400	347.07	339.96	13.26	7.11	96.25	97.95	97.09
	2021	400	344.78	343.57	27.73	1.21	92.53	99.65	95.96

森林变化 forest change	真实变化面积/ hm² true change area	预测变化面积/ hm² predicted change area	S_TP/ hm²	S_FP/ hm²	S_FN/ hm²	精确率/% precision	召回率/% recall	F₁分数/% F₁ score
森林减少 forest reduction	369.46	386.56	286.60	99.96	82.85	74.14	77.57	75.82
森林增加 forest increase	84.04	92.12	64.26	27.86	19.78	69.76	76.46	72.96
合计变化 total change	453.50	478.68	350.86	127.82	102.63	73.30	77.37	75.28

基于全卷积神经网络和低分辨率标签的森林变化检测研究

Research on forest change detection based on fully convolutional network and low resolution label

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 37

相关文章 5

编辑推荐

Metrics

本文评价

作者

读者

审稿

[1]	何乃磊, 张金生, 林文树. 基于深度学习多目标检测技术的林火图像识别研究[J]. 南京林业大学学报（自然科学版）, 2024, 48(3): 207-218.
[2]	牛弘健, 刘文萍, 陈日强, 宗世祥, 骆有庆. 基于Resnet的林地无人机图像去雾改进算法[J]. 南京林业大学学报（自然科学版）, 2024, 48(2): 175-18.
[3]	尹显明, 棘玉, 张日清, 莫登奎, 彭邵锋, 韦维. 深度学习在基于叶片的油茶品种识别中的研究[J]. 南京林业大学学报（自然科学版）, 2023, 47(3): 29-36.
[4]	徐姗姗, 吕净妍, 陈芳媛. 深度卷积神经网络的遥感植被检测方法[J]. 南京林业大学学报（自然科学版）, 2022, 46(4): 185-193.
[5]	刘嘉政, 王雪峰, 王甜. 基于深度学习的树种图像自动识别[J]. 南京林业大学学报（自然科学版）, 2020, 44(1): 138-144.