基于Sentinel⁃2影像与MaxEnt模型识别云南华坪县芒果种植区

doi:10.13304/j.nykjdb.2021.0409

摘要/Abstract

摘要：

云南省华坪县作为我国芒果的主要产地之一，近年来大力发展芒果种植业以及周边产业，促进了当地农业与旅游业的快速发展。为了更加科学高效地在空间尺度上对芒果种植区进行规划与管理，采用多时相Sentinel-2遥感影像构建多种植被指数，结合辅助地形因子与已挂果投产的芒果种植区野外调查点数据，通过MaxEnt模型对华坪县已投产芒果种植区进行分类识别，最后根据不同的阈值对预测结果进行二值化分类与精度评价。结果表明：二值化分类精度最高的阈值规则为10 percentile training presence，对应的阈值为0.257，分类总体精度为93.72%；在该阈值规则下估算已挂果投产的芒果种植区面积约为1.07万hm²，与研究时段内华坪县已有挂果投产的芒果种植区面积1.13万~1.20万hm²相近。因此，所选取的植被指数与地形因子组合在利用MaxEnt模型进行已投产芒果种植区的识别应用中取得了较好的效果，能为其他地区类似的研究应用提供借鉴，同时能为芒果种植业的发展与规划提供数据参考与决策支持。

关键词: Sentinel?2影像, MaxEnt模型, 芒果, 遥感识别

Abstract:

Huaping county of Yunnan province is one of the main mangos producing areas in China， and has vigorously developed mango planting and peripheral industries in recent years， which has promoted the rapid development of local agriculture and tourism. In order to plan and control manage mango planting areas in a more scientific and efficient way on the spatial scale， this study used multi-temporal sentinel-2 remote sensing images to build a variety of vegetation indexes， combined with the auxiliary terrain factors and the point data of mango planting areas that had been put into production obtained from field survey， and MaxEnt model was used to classify and identify the mango planting areas in Huaping county. Finally， binarization classification and accuracy evaluation of the predicted results were carried out according to different thresholds. The results showed that the highest threshold rule of binarization classification accuracy was 10 percentile training presence， the corresponding threshold was 0.257， and the overall classification accuracy was 93.72%. Under this threshold， the estimated area of mango planting area putting into production was about 1.07×10⁴hm²， which was close to （1.13~1.20）×10⁴ hm² of mango planting area putting into production in Huaping county during the research period. Therefore， the combination of vegetation indexes and terrain factors selected in this study achieved good results in the application of identification of mango planting areas by MaxEnt model， which provided some references for similar research and application in other areas， and provided some data references and decision supports for the development and planning of mango planting industry.

Key words: sentinel?2 images, MaxEnt model, mango, remote sensing recognition

中图分类号:

S126

赵管乐, 刘勤, 彭培好. 基于Sentinel⁃2影像与MaxEnt模型识别云南华坪县芒果种植区[J]. 中国农业科技导报, 2022, 24(3): 111-119.

Guanyue ZHAO, Qin LIU, Peihao PENG. Identification of Mango Planting Regions in Huaping County， Yunnan Province Based on Sentinel⁃2 Images and MaxEnt Model[J]. Journal of Agricultural Science and Technology, 2022, 24(3): 111-119.

图/表 7

参考文献 21

1	任传帅,叶回春,崔贝,等.基于面向对象分类的芒果林遥感提取方法研究[J].资源科学,2017,39(8):1584⁃1591.
	REN C S, YE H C, CUI B, et al.. Acreage estimation of mango orchards using object-oriented classification and remote sensing [J]. Resour. Sci., 2017,39(8):1584⁃1591.
2	李章成,王昕,李宗南,等.基于多时相高分一号影像水稻监测精度评价与修正——以德阳地区为例[J].西南农业学报,2016,29(10):2432⁃2435, 2505.
	LI Z C, WANG X, LI Z N, et al.. Classification accuracy evaluation and rejustment of monitoring rice planting area by multi-temporal GF-1 satellite image: taken Deyang district as example [J]. Southwest China J. Agric. Sci., 2016,29(10):2432⁃2435, 2505.
3	杨其坡.基于MaxEnt模型和多时相Sentinel⁃2影像的花椒识别研究[D].重庆:西南大学,2019.
	YANG Q P. Research on pepper identification based on maximum entropy model and multi-temporal sentinel-2 images [D]. Chongqing: Southwest University, 2019.
4	KOBAYASHI N, TANI H, WANG X F, et al.. Crop classification using spectral indices derived from Sentinel-2A imagery [J]. J. Inform. Telecommun., 2020,4(1):67⁃90.
5	PHILLIPS S J, ANDERSON R P, SCHAPIRE R E. Maximum entropy modelling of species geographic distributions [J]. Ecol. Model., 2006,190(3⁃4):231⁃259.
6	邢丁亮,郝占庆.最大熵原理及其在生态学研究中的应用[J].生物多样性,2011,19(3):295⁃302.
	XING D L, HAO Z Q. The principle of maximum entropy and its applications in ecology [J]. Biodiv. Sci., 2011,19(3):295⁃302.
7	PHILLIPS S J, DUDIK M, SCHAPIRE R E. A maximum entropy approach to species distribution modelling [C]// ICML. Proceedings of Association for Computing Machinery. Alberta, Canada, 2004: 655⁃662.
8	GUO Q, Li W, LIU Y, et al.. Predicting potential distributions of geographic events using one-class data: concepts and methods [J]. Int. J. Geographical Inform. Sci., 2011,25(10):1697⁃1715.
9	田庆久,闵祥军.植被指数研究进展[J].地球科学进展,1998(4):10⁃16.
	TIAN Q J, MIN X J. Advances in Study on Vegetation Indices [J]. Adva. Earth Sci., 1998(4):10⁃16.
10	TUCKER C J. Red and photographic infrared linear combinations for monitoring vegetation [J]. Remote Sens. Environ., 1979,8(2):127⁃150.
11	JORDAN C F. Derivation of leaf‐area index from quality of light on the forest floor [J]. Ecology, 1969,50(4): 663⁃666.
12	PEARSON R L, MILLER L D. Remote mapping of standing crop biomass for estimation of the productivity of the short grass prairie [C]// Proceedings of 8th International Symptom Remote Sensing Environment. Ann Arbor, Michigan USA, 1972:1355.
13	HUETE A R. A soil-adjusted vegetation index (SAVI) [J]. Remote Sens. Environ., 1988,25(3):295⁃309.
14	ClEVERS J G P W. The application of a weighted infrared-red vegetation index for estimating leaf Area index by correcting for soil moisture [J]. Remote Sens. Environ., 1989,29(1):25⁃37.
15	BLACKBURN G A. Quantifying chlorophylls and carotenoids at leaf and canopy scales: an evaluation of some hyperspectral approaches [J]. 1998,66(3):273⁃285.
16	刘佳,王利民,滕飞,等. RapidEye卫星红边波段对农作物面积提取精度的影响[J].农业工程学报,2016,32(13):140⁃148.
	LIU J, WANG L M, TENG F, et al.. Impact of red-edge waveband of RapidEye satellite on estimation accuracy of crop planting area [J]. Trans. Chin. Soc. Agric. Eng., 2016,32(13):140⁃148.
17	吴晓婷,闫德勤.数据降维方法分析与研究[J].计算机应用研究,2009,26(8):2832⁃2835.
	WU X T, YAN D Q. Analysis and research on method of data dimensionality reduction [J]. Appl. Res. Comp., 2009,26(8):2832⁃2835.
18	赵祎.浅析华坪生产优质晚熟绿色芒果的自然环境要素[J].中国热带农业,2013(5):37⁃39.
19	付发群.基于Sentinel-1A与Sentinel-2A数据的城市绿地提取研究[D].浙江金华:浙江师范大学,2020.
	FU F Q. Study of urban green space extraction based on Sentinel-1A and Sentinel-2A data [D]. Zhejiang Jinhua: Zhejiang Normal University, 2020.
20	王克晓,周蕊,欧毅,等.基于光学与雷达遥感的山地油菜分布提取研究[J].西南大学学报(自然科学版),2021,43(6):139⁃146.
	WANG K X, ZHOU R, OU Y, et al.. Research on extraction of information about rape distribution in hilly regions based on optical and radar images [J]. J. Southwest Univ. (Nat. Sci.), 2021,43(6):139⁃146.
21	赵娟. 基于深度神经网络模型的芒果种植区遥感提取[D].成都:西南交通大学,2019.
	ZHAO J. Extraction of cultivation area of mango from remotely sensed imageries based on deep network models-a case study in Renhe district of Panzhihua city [D]. Chengdu: Southwest Jiaotong University, 2019.

环境变量 Environmental variable	相对贡献率 Percent contribution/%	累计贡献率 Cumulative percent contribution/%
DEM	27.8	27.8
20191216DVI	15.0	42.8
20191216WDVI	10.6	53.4
20200817PCA1	9.8	63.2
20200608WDVI	9.0	72.2
20200608SAVI	6.2	78.4
20200315DVI	5.6	84.0
坡向Aspect	4.0	88.0
20200817WDVI	3.3	91.3
20200608PCA1	2.6	93.9
20200315PSSRa	1.3	95.2
20200315WDVI	1.1	96.3
20191216PSSRa	0.7	97.0
坡度Slope	0.6	97.6
20191216NDVI	0.5	98.1
20200315PCA1	0.4	98.5
20191216PCA1	0.4	98.9
20200817SAVI	0.2	99.1
20200608NDVI	0.2	99.3
20200608PSSRa	0.2	99.5
20200817PSSRa	0.2	99.7
20200608DVI	0.1	99.8
20200817DVI	0.1	99.9
20200817RVI	0.1	100.0

环境变量 Environmental variable	相对贡献率 Percent contribution/%	累计贡献率 Cumulative percent contribution/%
DEM	27.8	27.8
20191216DVI	15.0	42.8
20191216WDVI	10.6	53.4
20200817PCA1	9.8	63.2
20200608WDVI	9.0	72.2
20200608SAVI	6.2	78.4
20200315DVI	5.6	84.0
坡向Aspect	4.0	88.0
20200817WDVI	3.3	91.3
20200608PCA1	2.6	93.9
20200315PSSRa	1.3	95.2
20200315WDVI	1.1	96.3
20191216PSSRa	0.7	97.0
坡度Slope	0.6	97.6
20191216NDVI	0.5	98.1
20200315PCA1	0.4	98.5
20191216PCA1	0.4	98.9
20200817SAVI	0.2	99.1
20200608NDVI	0.2	99.3
20200608PSSRa	0.2	99.5
20200817PSSRa	0.2	99.7
20200608DVI	0.1	99.8
20200817DVI	0.1	99.9
20200817RVI	0.1	100.0

阈值规则 Threshold rule	描述 Description	阈值 Threshold	总体精度 Overall accuracy/%	Kappa系数 Kappa coefficient
EETOD	使阈值分布和原始分布的熵相等 Equate entropy of thresholded and original distributions	0.118	89.57	0.73
ETESS	相同的检测灵敏度和特异性 Equal test sensitivity and specificity	0.193	92.93	0.81
ETRSS	相同的训练敏感性和特异性 Equal training sensitivity and specificity	0.201	92.97	0.81
FCV	固定累积值10 Fixed cumulative value 10	0.132	90.18	0.75
MTESS	最大的测试灵敏度和特异性 Maximum test sensitivity plus specificity	0.140	91.39	0.77
MTRSS	最大的训练灵敏度和特异性 Maximum training sensitivity plus specificity	0.169	91.57	0.78
PTP	10%训练量 10 percentile training presence	0.257	93.72	0.82

阈值规则 Threshold rule	描述 Description	阈值 Threshold	总体精度 Overall accuracy/%	Kappa系数 Kappa coefficient
EETOD	使阈值分布和原始分布的熵相等 Equate entropy of thresholded and original distributions	0.118	89.57	0.73
ETESS	相同的检测灵敏度和特异性 Equal test sensitivity and specificity	0.193	92.93	0.81
ETRSS	相同的训练敏感性和特异性 Equal training sensitivity and specificity	0.201	92.97	0.81
FCV	固定累积值10 Fixed cumulative value 10	0.132	90.18	0.75
MTESS	最大的测试灵敏度和特异性 Maximum test sensitivity plus specificity	0.140	91.39	0.77
MTRSS	最大的训练灵敏度和特异性 Maximum training sensitivity plus specificity	0.169	91.57	0.78
PTP	10%训练量 10 percentile training presence	0.257	93.72	0.82

[1]	马春英1,张永升2,侯俊玲3,王文全3,4. 移栽密度对知母生长性状及药用活性成分含量的影响[J]. 中国农业科技导报, 2017, 19(6): 106-110.
[2]	杨杨1,2,3,申琳2,生吉萍2,4,王凤忠1,王锋1,范蓓1*. 外源NO对采后芒果果实低温胁迫后H2O2代谢的影响[J]. , 2013, 15(2): 131-136.