机载多光谱影像语义分割模型在农田防护林提取中的应用

doi:10.12082/dqxxkx.2020.200086

地球信息科学学报 ›› 2020, Vol. 22 ›› Issue (8): 1702-1713.doi: 10.12082/dqxxkx.2020.200086

机载多光谱影像语义分割模型在农田防护林提取中的应用

王学文¹^,²(), 赵庆展¹^,²^,^*(), 韩峰¹^,², 马永建¹^,², 龙翔²^,³, 江萍²^,³

1.石河子大学信息科学与技术学院,石河子 832003
2.兵团空间信息工程技术研究中心,石河子 832003
3.石河子大学机械电气工程学院,石河子 832003

收稿日期:2020-02-23 修回日期:2020-03-25 出版日期:2020-08-25 发布日期:2020-10-25
通讯作者: 赵庆展 E-mail:wangxuewen@stu.shzu.edu.cn;zqz_inf@shzu.edu.cn
作者简介:王学文（1996— ）,男,江苏南通人,硕士生,主要从事农业信息化技术及应用、深度学习与遥感影像处理研究。 E-mail:wangxuewen@stu.shzu.edu.cn
基金资助:
新疆生产建设兵团科技计划项目(2017DB005);兵团空间信息工程技术研究中心创建项目(2016BA001);中央引导地方科技发展专项资金项目(201610011)

Application of Airborne Multispectral Image Semantic Segmentation Model in Farmland Shelterbelt Extraction

WANG Xuewen¹^,²(), ZHAO Qingzhan¹^,²^,^*(), HAN Feng¹^,², MA Yongjian¹^,², LONG Xiang²^,³, JIANG Ping²^,³

1. College of Information Science & Technology, Shihezi University, Shihezi 832003, China
2. Geospatial Information Engineering Research Center, Xinjiang Production and Construction Corps, Shihezi 832003, China
3. College of Mechanical and Electrical Engineering, Shihezi University, Shihezi 832003, China

Received:2020-02-23 Revised:2020-03-25 Online:2020-08-25 Published:2020-10-25
Contact: ZHAO Qingzhan E-mail:wangxuewen@stu.shzu.edu.cn;zqz_inf@shzu.edu.cn
Supported by:
Xinjiang Production and Construction Corps Science and Technology Project(2017DB005);Geospatial Information Engineering Research Center To Create, Xin-jiang Production and Construction Corps(2016BA001);The Central Government Directs Local Science and Technology Development Special Funds(201610011)

摘要/Abstract

摘要：

农田防护林是农田生态系统的屏障,其健康状况的监测与评估在我国北方农田林网管理中尤为重要。本文以新疆生产建设兵团第三师51团为研究区,使用复合翼无人机CW-20搭载Micro MCA12 Snap多光谱相机获取农田防护林的多光谱影像,经辐射校正、裁剪等预处理,通过优选有效特征和模型比较,提出农田防护林提取的有效方法。首先,基于原始12波段,依据相关性系数矩阵和最佳指数因子（Optimum Index Factor,OIF）选取最优3波段和植被指数特征进行组合,构建8种农田防护林提取方案;然后,通过建立语义分割Deeplabv3+模型进行精度评价,得到最优3波段组合6（波长710 nm）、8（波长800 nm）、 11（波长900 nm）波段为最佳特征组合;最后,以最优3波段为基础,将Deeplabv3+模型与U-Net、ENVINet5模型进行对比分析。结果表明：Deeplabv3+模型能够更深层次的挖掘光谱中潜在的信息,相比其他模型,能够较好地处理正负样本不均衡问题,获得最高MIoU值85.54%,比U-Net、ENVINet5的MIoU值则分别高出21.21%、27.19%。该研究结果可为基于多光谱遥感影像的语义分割在农田防护林提取及健康状况监测的应用提供借鉴和参考。

关键词: 无人机, 多光谱影像, 农田防护林提取, 最佳指数因子, 植被指数, 特征选择, 语义分割, Deeplabv3+

Abstract:

Farmland shelterbelt is the barrier of a farmland ecosystem. Monitoring and assessing the health condition of farmland shelterbelt are significant for farmland forests management in North China. This paper selects the farmland shelterbelt in the 3^th Division of Xinjiang Production and Construction Crops as the experimental example. In July and August of 2019, we used the CW-20 fixed-wing drone which was equipped with a SONY-A7RII camera and a Micro MCA12 Snap camera to capture visible light images and multispectral images. Four targets on the ground with different reflectance (3%, 22%, 48%, and 64%) were set up for radiometric correction. we collected field data by Aowei software. Optimal bands were selected base on the Optimum Index Factor (OIF) and correlation coefficient. By integrating appropriate vegetation indices, eight different extraction schemes were constructed to select an effective method for final farmland shelterbelt extraction. In this study, when the flight mission was completed, we collated the exported POS and image data correspondingly. High resolution visible light images with 0.08 m resolution and high resolution multispectral images with 0.149 m resolution were first mosaiced using Pix4Dmapper. Four ground targets were then used for radiometric correction of mosaiced images using linear fitting model. The final experimental data was extracted using clipping method. Because each multispectral image was displayed in gray scale, which was not easy for direct data interpretation. High-resolution visible light images and multispectral images were registered together through ArcGIS software. We labeled 8030 samples based on visible light images through ENVI 5.5 software. We separated the training dataset and verification dataset according to the ratio of 4:1. Specifically, 112 512 pixel×512 pixel slices were taken as the training dataset, and 28 512 pixel×512 pixel slices were taken as the verification dataset. Based on eight extraction schemes, model accuracy of Deeplabv3+semantic segmentation model was evaluated. Three optimal bands that had the best model performance were selected including band 6 (wavelength 710 nm), band 8 (wavelength 800 nm), and band 11 (wavelength 900 nm) based on OIF. Based on these three optimal bands, we further compared the Deeplabv3+ model with U-Net and ENVINet5 model. Our results show that Deeplabv3 + model can capture the potential information in the images at a deeper level with fewer parameters. The MIoU derived from Deeplabv3+, U-Net, and ENVINet5 was 85.54%, 21.21%, and 27.19%, respectively. Also, Deeplabv3 + model solved the problem of sample unbalance. Our results provide a reference for the application of semantic segmentation using multispectral remote sensing data in farmland areas.

Key words: The Unmanned Aerial Vehicle (UAV), multispectral image, farmland shelterbelt extraction, optimum index factor, vegetation index, feature combination, semantic segmentation, Deeplabv3+

王学文, 赵庆展, 韩峰, 马永建, 龙翔, 江萍. 机载多光谱影像语义分割模型在农田防护林提取中的应用[J]. 地球信息科学学报, 2020, 22(8): 1702-1713.DOI:10.12082/dqxxkx.2020.200086

WANG Xuewen, ZHAO Qingzhan, HAN Feng, MA Yongjian, LONG Xiang, JIANG Ping. Application of Airborne Multispectral Image Semantic Segmentation Model in Farmland Shelterbelt Extraction[J]. Journal of Geo-information Science, 2020, 22(8): 1702-1713.DOI:10.12082/dqxxkx.2020.200086

图/表 14

图1

表1

图2

图3

表2

植被指数及计算公式"

指数	中文名称	英文名称	对应Micro MCA12 Snap波段	计算公式及编号
NDVI^[25]	归一化植被指数	Normalized difference vegetation index	B8、B5	$NDVI = ρ 800 - ρ 656 ρ 800 + ρ 656$	（3）
NDVIre^[26]	红边归一化差值植被指数	Red-edge normalized difference vegetation index	B8、B7、B8、B6	$NDVIre = ρ 800 - ρ 760710 ρ 800 + ρ 760710$	（4）
MTCI^[27]	地面叶绿素指数	MERIS Terrestrial Chlorophyll Index	B7、B6、B5	$MTCI = ρ 760 - ρ 710 ρ 710 - ρ 656$	（5）
NDRE^[28]	归一化差值红边指数	Normalized difference Red Edge index	B7、B6	$NDRE = ρ 760 - ρ 710 ρ 760 + ρ 710$	（6）

表2

表3

图4

表4

表5

图5

表6

图6

图7

图8

参考文献 33

[1]	Vigiak O, Sterk G, Warren A, et al. Spatial modeling of wind speed around windbreaks[J]. Catena, 2003,52(3-4):273-288.
[2]	Brandle J R, Hodges L, Zhou X H. Windbreaks in North American agricultural systems[J]. Agroforestry Systems, 2004,61(1-3):65-78.
[3]	陈伟. 基于高标准农田防护林建设若干问题的探讨[J]. 现代园艺, 2019(14):215-216.
	[ Chen W. Discussion on some problems of farmland shelterbelt construction based on high standard[J]. Xiandai Horticulture, 2019(14):215-216. ]
[4]	王菊兰. 金塔县农田防护林体系现状与建议[J]. 甘肃林业科技, 2019(1):34-37.
	[ Wang J L. Current situation and suggestions of farmland shelterbelt system in Jinta County[J]. Journal of Gansu Forestry Science and Technology, 2019(1):34-37. ]
[5]	Wessel M, Brandmeier M, Tiede D. Evaluation of different machine learning algorithms for scalable classification of tree types and tree species based on Sentinel-2 data[J]. Remote Sensing, 2018,10(9):1419.
[6]	Anees A, Aryal J. A statistical framework for near-real time detection of beetle infestation in pine forests using MODIS data[J]. IEEE Geoscience and remote sensing letters, 2014,11(10):1717-1721.
[7]	Torresan C, Wallace L. Forestry applications of UAVs in Europe: A review[J]. International Journal of Remote Sensing, 2017,38(8-10):2427-2447.
[8]	Kort J, Poppy L, Gordon A, et al. Quantification of shelterbelt characteristics using high-resolution imagery[J]. Agriculture Ecosystems & Environment, 2009,131(1):111-117.
[9]	Shang W T, Gao Z, Jiang X P, et al. The extraction of coastal windbreak forest information based on UAV remote sensing images[C]// Remote Sensing and Modeling of Ecosystems for Sustainability XIV. Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, 2017.
[10]	Johansen K, Raharjo T, McCabe M F. Using multi-spectral UAV imagery to extract tree crop structural properties and assess pruning effects[J]. Remote Sensing, 2018,10(6):854.
[11]	Hamdi Z M, Brandmeier M, Straub C. Forest damage assessment using deep learning on high resolution remote sensing data[J]. Remote Sensing, 2019,11(17):1976.
[12]	赵庆展, 刘伟, 尹小君, 等. 基于无人机多光谱影像特征的最佳波段组合研究[J]. 农业机械学报, 2016,47(3):242-248,291.
	[ Zhao Q Z, Liu W, Yin X J, et al. Selection of optimum bands combination based on multispectral images of UAV[J]. Transactions of the Chinese Society for Agricultural Machinery, 2016,47(3):242-248,291. ]
[13]	顾铮鸣, 金晓斌, 杨晓艳, 等. 基于无人机遥感影像监测土地整治项目道路沟渠利用情况[J]. 农业工程学报, 2018,34(23):85-93.
	[ Gu Z M, Jin X B, Yang X Y, et al. Monitoring roads and canals utilization condition for land consolidation project based on UAV remote sensing image[J]. Transactions of the Chinese Society of Agricultural Engineering, 2018,34(23):85-93. ]
[14]	张磊, 宫兆宁, 王启为, 等. Sentinel-2影像多特征优选的黄河三角洲湿地信息提取[J]. 遥感学报, 2019,23(2):313-326.
	[ Zhang L, Gong Z N, Wang Q W, et al. Wetland mapping of Yellow River Delta wetlands based on multifeature optimization of Sentinel-2 images[J]. Journal of Remote Sensing, 2019,23(2):313-326. ]
[15]	李冠东, 张春菊, 高飞, 等. 双卷积池化结构的3D-CNN高光谱遥感影像分类方法[J]. 中国图象图形学报, 2019,24(4):639-654.
	[ Li G D, Zhang C J, Gao F, et al. Double convolution pool-structured 3D-CNN for hyperspectral remote sensing image classification[J]. Journal of Image and Graphics, 2019,24(4):639-654. ]
[16]	杨建宇, 周振旭, 杜贞容, 等. 基于SegNet语义模型的高分辨率遥感影像农村建设用地提取[J]. 农业工程学报, 2019,35(5):251-258.
	[ Yang J Y, Zhou Z X, Du Z R, et al. Rural construction land extraction from high spatial resolution remote sensing image based on SegNet semantic segmentation model[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019,35(5):251-258. ]
[17]	Yao X, Yang H, Wu Y, et al. Land use classification of the deep convolutional neural network method reducing the loss of spatial features[J]. Sensors, 2019,19(12):2792.
[18]	Krestenitis M, Orfanidis G, Ioannidis K, et al. Oil spill identification from satellite images using deep neural networks[J]. Remote Sensing, 2019,11(15):1762.
[19]	Morales G, Kemper G, Sevillano G, et al. Automatic segmentation of mauritia flexuosa in Unmanned Aerial Vehicle (UAV) Imagery using deep learning[J]. Forests, 2018,9(12):736.
[20]	贾银江. 无人机遥感图像拼接关键技术研究[D]. 黑龙江:东北农业大学, 2016.
	[ Jia Y J. Research on Key Technologies of UAV remote sensing image mosaic[D]. Heilongjiang: Northeast Agricultural University, 2016. ]
[21]	相涛, 栾元重, 许章平, 等. 基于Pix4Dmapper的无人机低空摄影测量数据处理[J]. 测绘与空间地理信息, 2019,42(3):75-78.
	[ Xiang T, Luan Y Z, Xu Z P, et al. Unmanned aerial vehicle with low-altitude aerial photogrammetry data processing flow and precision analysis based on pix4Dmapper[J]. Geomatics & spatial information technology, 2019,42(3):75-78. ]
[22]	Ji W, Wang Z Y, Wen Q. A fusion method of hyperspectral image based on spectral high fidelity applied in spectrum retrieval of vegetation species[J]. Multispectral hyperspectral and ultraspectral remote sensing technology techniques and applications, 2014(9263):1-8.
[23]	Chavez P S, Berlin G L, Sowers L B. Statistical-method for selecting Landsat mss rations[J]. Journal of applied photographic engineering, 1982,8(1):23-30.
[24]	郭文婷, 张晓丽. 基于Sentinel-2时序多特征的植被分类[J]. 浙江农林大学学报, 2019,36(5):849-856.
	[ Guo W T, Zhang X L. Vegetation classification based on a multi-feature Sentinel-2 time series[J]. Journal of Zhejiang A &F University, 2019,36(5):849-856. ]
[25]	孙庆龄, 李宝林, 许丽丽, 等. 2000-2013年三江源植被NDVI变化趋势及影响因素分析[J]. 地球信息科学学报, 2016,18(12):1707-1716.
	[ Sun Q L, Li B L, Xu L L, et al. Analysis of NDVI change trend and its impact factors in the Three-River Headwater Region from 2000 to 2013[J]. Journal of Geo-information Science, 2016,18(12):1707-1716. ]
[26]	Shang J, Liu J, Ma B, et al. Mapping spatial variability of crop growth conditions using RapidEye data in Northern Ontario Canada[J]. Remote Sensing of Environment, 2015(168):113-125.
[27]	Dash J, Curran P J. The MERIS terrestrial chlorophyll index[J]. International Journal of Remote Sensing, 2004,25(23):5403-5413.
[28]	Gitelson A, Merzlyak M N. Spectral reflectance changes associated with autumn senescence of aesculus hippocastanum L and Acer platanoides L leaves spectral features and relation to Chlorophyll estimation[J]. Journal of Plant Physiology, 1994,143(3):286-292.
[29]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016.
	[ Zhou Z H. Machine learning[M]. Beijing: Tsinghua University Pubilshing, 2016. ]
[30]	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, 2015: 234-241.
[31]	许慧敏. 基于深度学习U-Net模型的高分辨率遥感影像分类方法研究[D]. 成都:西南交通大学, 2018.
	[ Xu H M. Research on high resolution remote sensing image classification method based on deep learning U-Net model[D]. Chendu: Southwest Jiaotong University, 2018. ]
[32]	Environmental Systems Research Institut, ENVI deep learning[ED/OL]. https://www. harrisgeospatial.com/SoftwareTechnology/ENVI-Deep-Learning, 2019-04-01/2020-03-09.
[33]	贺浩, 王仕成, 杨东方, 等. 基于Encoder- Decoder网络的遥感影像道路提取方法[J]. 测绘学报, 2019,48(3):330-338.
	[ He H, Wang S C, Yang D F, et al. An road extraction method for remote sensing image based on Encoder-Decoder network[J]. Acta Geodaetica et Cartographica Sinica, 2019,48(3):330-338. ]

实验方案	特征组合
1	多光谱原始12波段
2	最佳指数因子选出3波段
3	最优3波段+NDVI
4	最优3波段+NDVIre
5	最优3波段+MTCI
6	最优3波段+NDRE
7	NDVI+NDVIre+MTCI+NDRE
8	最优3波段+NDVI+NDVIre+MTCI+NDRE

波段	波段
波段	1	2	3	4	5	6	7	8	9	10	11	12
1	1.000
2	0.982	1.000
3	0.960	0.966	1.000
4	0.968	0.973	0.963	1.000
5	0.958	0.962	0.940	0.991	1.000
6	0.827	0.850	0.922	0.862	0.831	1.000
7	-0.330	-0.302	-0.192	-0.357	-0.395	0.087	1.000
8	-0.330	-0.300	-0.188	-0.357	-0.398	0.099	0.966	1.000
9	-0.345	-0.315	-0.204	-0.372	-0.410	0.081	0.964	0.982	1.000
10	-0.339	-0.309	-0.198	-0.364	-0.402	0.089	0.967	0.981	0.971	1.000
11	-0.335	-0.308	-0.199	-0.362	-0.397	0.080	0.965	0.952	0.947	0.979	1.000
12	-0.261	-0.239	-0.119	-0.281	-0.319	0.169	0.931	0.943	0.944	0.960	0.957	1.000

序号	波段组合	OIF
1	6、8、11	25.756
2	6、8、10	24.962
3	6、10、11	24.881
4	6、9、11	24.632
5	6、7、8	24.136

地物类型	指标	方案1	方案2	方案3	方案4	方案5	方案6	方案7	方案8
农田防护林	F1值	0.90	0.92	0.88	0.91	0.79	0.90	0.71	0.76
	查准率precision/%	84.83	88.40	81.94	86.28	67.57	86.06	39.13	62.29
	查全率recall/%	95.09	95.42	95.49	95.62	95.57	94.84	94.73	96.10
	IoU/%	73.66	79.21	69.41	75.87	51.02	75.53	24.32	45.23
其他地物	F1值	0.91	0.92	0.90	0.91	0.85	0.91	0.81	0.83
	查准率precision/%	95.62	95.76	96.13	96.05	96.87	95.32	98.44	97.46
	查全率recall/%	86.31	89.20	84.84	87.50	74.92	87.24	69.30	72.10
	IoU/%	91.61	91.61	91.86	92.40	92.14	91.06	96.93	94.13
平均	MIoU/%	82.63	85.54	80.98	84.14	72.48	83.29	60.63	69.75

机载多光谱影像语义分割模型在农田防护林提取中的应用

Application of Airborne Multispectral Image Semantic Segmentation Model in Farmland Shelterbelt Extraction

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 14

参考文献 33

相关文章 15

Metrics

本文评价

推荐阅读 0

波段	波长/nm	波宽/nm	波段特征
1	470	10	区分植被和土壤岩石表层
2	515	10	可见光光谱中的绿波波峰
3	550	10	对水质浑浊变化敏感
4	610	10	植被光谱反射曲线中红光初始波段
5	656	10	重新归一化植被指数
6	710	10	植被光谱反射曲线的红边波段
7	760	10	植被光谱反射曲线的红边波段
8	800	10	归一化植被指数
9	830	10	区别植被种类
10	860	10	与植物总叶绿素含量显著相关
11	900	20	计算特定作物水分敏感指数
12	950	40	计算水波段指数

[1]	王艳杰, 王卷乐, 魏海硕, Altansukh Ochir, Davaadorj Davaasuren, Sonomdagva Chonokhuu. 基于稀疏样点的蒙古国产草量估算方法研究[J]. 地球信息科学学报, 2020, 22(9): 1814-1822.
[2]	陈昂, 杨秀春, 徐斌, 金云翔, 张文博, 郭剑, 邢晓语, 杨东. 基于面向对象与深度学习的榆树疏林识别方法研究[J]. 地球信息科学学报, 2020, 22(9): 1897-1909.
[3]	陈如如, 胡中民, 李胜功, 郭群. 不同数据源归一化植被指数在中国北方草原区的应用比较[J]. 地球信息科学学报, 2020, 22(9): 1910-1919.
[4]	蒋世豪, 江洪, 陈慧. 基于SEVI的复杂地形山区植被FPAR遥感反演与地形效应评估[J]. 地球信息科学学报, 2020, 22(8): 1725-1734.
[5]	蔡顺, 耿豪鹏, 郑炜珊, 潘保田. 基于傅里叶变换的谷间距特征信息提取及其影响因素研究[J]. 地球信息科学学报, 2020, 22(3): 399-409.
[6]	曹泽涛, 方子东, 姚瑾, 熊礼阳. 基于随机森林的黄土地貌分类研究[J]. 地球信息科学学报, 2020, 22(3): 452-463.
[7]	张志慧, 刘雯, 李笑含, 朱靖轩, 张洪涛, 杨东, 徐超昊, 徐宪立. 基于无人机航摄影像的喀斯特地区裸岩信息提取及景观格局分析[J]. 地球信息科学学报, 2020, 22(12): 2436-2444.
[8]	许佳峰, 李云梅, 徐杰, 雷少华, 毕顺, 周玲. 黑臭水体水面阴影提取的自适应阈值算法研究[J]. 地球信息科学学报, 2020, 22(10): 1959-1970.
[9]	耿仁方,付波霖,蔡江涛,陈晓雨,蓝斐芜,余杭洺,李青逊. 基于无人机影像和面向对象随机森林算法的岩溶湿地植被识别方法研究[J]. 地球信息科学学报, 2019, 21(8): 1295-1306.
[10]	鹿明, 廖小罕, 岳焕印, 黄诗峰, 徐晨晨, 卢海英, 柏艺琴. 面向中国洪涝灾害应急监测的无人机空港布局[J]. 地球信息科学学报, 2019, 21(6): 854-864.
[11]	吴方明, 张淼, 吴炳方. 无人机影像的面向对象水稻种植面积快速提取[J]. 地球信息科学学报, 2019, 21(5): 789-798.
[12]	王森援, 蔡国榕, 王宗跃, 吴云东. 基于加权约束的单体建筑物点云表面重建算法[J]. 地球信息科学学报, 2019, 21(5): 654-662.
[13]	晏磊, 廖小罕, 周成虎, 樊邦奎, 龚健雅, 崔鹏, 郑玉权, 谭翔. 中国无人机遥感技术突破与产业发展综述[J]. 地球信息科学学报, 2019, 21(4): 476-495.
[14]	杨梦琳, 李道春, 万志强, 严德, 王耀坤. 中国面向遥感应用的垂直起降无人机现状和应用前景分析[J]. 地球信息科学学报, 2019, 21(4): 496-503.
[15]	黄耀欢, 李中华, 朱海涛. 作物胁迫无人机遥感监测研究评述[J]. 地球信息科学学报, 2019, 21(4): 512-523.