Effects of Different Sample Set Partition Strategies on Crop Remote Sensing Classification Accuracy

doi:10.15933/j.cnki.1004-3268.2024.06.016

Abstract

Abstract: The extraction accuracy of crop distribution has a profound impact on the subsequent inversion of farmland parameters and estimation of crop yield per unit area.In the process of crop classification and recognition,the accuracy and number of training samples are crucial to the final classification results.Aiming at the problem of small number of samples and uneven distribution,the crop classification sample data set was constructed by two ways of field identification and visual interpretation,and five sample data set construction schemes were designed:①all the measured sample points(70% training,30% verification)were used in the scheme;②all visual interpretation sample points were used(70% training,30% verification);③ the same proportion of training samples and verification samples were selected from the measured sample points and visual sample points respectively,and then the training sample set and verification sample set were constructed combinedly(70% training,30% verification);④the visual interpretation sample points were used as training samples,and the measured sample points were used as verification samples;⑤the same number of samples were selected from the visual sample points and the measured sample points to construct a sample set(70% training,30% verification).The accuracy of crop remote sensing classification was studied under different schemes.The results showed that except④,the overall accuracy of ① ② ③ ⑤ four sample data set partition schemes was more than 95%,and the classification results were good. Using visual interpretation to supplement sample points could effectively solve the problem of fewer sample points and uneven distribution.As the best classification scheme for crop recognition and extraction in the study area,scheme③had an overall accuracy of 97.6% and a Kappa coefficient of 0.970,and the accuracy of corn,rice and soybean was all more than 97%,indicating that the combination of training samples and validation samples selected from visual interpretation samples and measured samples to construct training and validation sample sets can not only improve the accuracy of classification results,but also improve the authenticity and accuracy of classification results.

Key words: Remote sensing, Crop classification, Visual interpretation, Sample set constraction

摘要： 农作物分布的提取精度对于后续的农田参数反演和作物单产估算等均具有深刻的影响，而农作物分类识别过程中，训练样本的准确性和数量对其最终分类结果的影响是至关重要的。针对样本较少且分布不均匀的问题，通过实地标识和目视解译2种方式构建农作物分类样本数据集，设计5种样本数据集构建方案：方案①全部采用实测样点（70%训练、30%验证）；方案②全部采用目视解译样点（70%训练、30%验证）；方案③实测样点与目视样点分别选取相同比例的训练样本与验证样本，再结合构建训练样本集与验证样本集（70%训练、30%验证）；方案④目视解译样点作为训练样本，实测样点作为验证样本；方案⑤目视样点与实测样点选取相同数量的样本进行结合构建样本集（70%训练、30%验证）。研究不同方案下的农作物遥感分类精度。结果表明，除方案④外，①②③⑤4种样本数据集划分方案的整体精度均在95%以上，分类结果较好，表明采用目视解译补充样本点，可以有效解决样本点较少和分布不均匀的情况。方案③作为研究区作物识别提取的最佳分类方案，总体精度达到97.6%，Kappa系数达到0.970，且玉米、水稻、大豆单类精度均超过97%，表明目视解译样点与实测样点分别选取训练样本与验证样本再结合构建训练样本集与验证样本集，不仅可以提升分类结果的精度，而且可以提高分类结果的真实性、准确性。

关键词: 遥感, 农作物分类, 目视解译, 样本集划分

CLC Number:

S127

LIU Yang, LI Qiangzi, DU Xin, WANG Hongyan, ZHANG Yuan, ZHANG Xiwang, SHEN Yunqi, ZHANG Sichen, YU Shiqi. Effects of Different Sample Set Partition Strategies on Crop Remote Sensing Classification Accuracy[J]. Journal of Henan Agricultural Sciences, 2024, 53(6): 144-153.

刘洋, 李强子, 杜鑫, 王红岩, 张源, 张喜旺, 沈云祺, 张思宸, 余仕奇. 不同样本集划分策略对农作物遥感分类精度的影响[J]. 河南农业科学, 2024, 53(6): 144-153.

References

［1］樊东东，李强子，王红岩，等.通过训练样本采样处理改善小宗作物遥感识别精度［J］.遥感学报，2019，23（4）：730‑742.
FAN D D，LI Q Z，WANG H Y，et al.Improvement in recognition accuracy of minority crops by resampling of

imbalanced training datasets of remote sensing［J］.Journal of Remote Sensing，2019，23（4）：730‑742.

［2］王娜，李强子，杜鑫，等.单变量特征选择的苏北地区主要农作物遥感识别［J］.遥感学报，2017，21（4）：519‑530.

WANG N，LI Q Z，DU X，et al. Identification of main crops based on the univariate feature selection in Subei［J］.Journal of Remote Sensing，2017，21（4）：519‑530.

［3］张健康，程彦培，张发旺，等.基于多时相遥感影像的作物种植信息提取［J］.农业工程学报，2012，28（2）：134‑141.

ZHANG J K，CHENG Y P，ZHANG F W，et al.Crops planting information extraction based on multi‑temporal remote sensing images［J］.Transactions of the Chinese Society of Agricultural Engineering，2012，28（2）：134‑141.

［4］朱文泉，潘耀忠，龙中华，等.基于GIS和RS的区域陆地植被NPP估算：以中国内蒙古为例［J］.遥感学报，2005，9（3）：300‑307.

ZHU W Q，PAN Y Z，LONG Z H，et al.Estimating net primary productivity of terrestrial vegetation based on GIS and RS：A case study in inner Mongolia，China［J］.Journal of Remote Sensing，2005，9（3）：300‑307.

［5］张爽，刘雪华，靳强.决策树学习方法应用于生境景观分类［J］.清华大学学报（自然科学版），2006，46（9）：1564‑1567.

ZHANG S，LIU X H，JIN Q.Decision tree learning for habitat landscape classification［J］.Journal of Tsinghua University（Science and Technology），2006，46（9）：1564‑1567.

［6］许淇.基于随机森林算法的遥感分类方法优化研究：以青铜峡灌区为例［D］.太谷：山西农业大学，2019.

XU Q.Research on optimization of remote sensing classification methods based on random forest algorithm：A case of Qingtongxia irrigation district［D］.Taigu：Shanxi Agricultural University，2019.

［7］沈峰，石璞，井逢瑞.基于Sentinel-2影像的吉林省中部黑土区保护性耕作范围的遥感识别［J］.世界地质，2022，41（4）：873‑881.

SHEN F，SHI P，JING F R.Remote sensing identification of conservation tillage extent in black soil region in central Jilin Province based on Sentinel‑2 image［J］.World Geology，2022，41（4）：873‑881.

［8］尚明，马杰，李悦，等.Landsat和GF数据面向对象土地覆盖分类研究［J/OL］.自然资源遥感，2023：1‑8［2023‑12‑26］.https：//kns.cnki.net/kcms/detail/10.1759.P.20231225.1527.010.html.

SHANG M，MA J，LI Y，et al.Research on object‑oriented land cover classification of Landsat and GF data［J/OL］.Remote Sensing for Natural Resources，2023：1‑8［2023‑12‑26］.https：//kns.cnki.net/kcms/detail/10.1759.P.20231225.1527.010.html.

［9］刘杰，刘吉凯，安晶晶，等.基于时序Landsat 8 OLI多特征与随机森林算法的作物精细分类研究［J］.干旱地区农业研究，2020，38（3）：281‑288.

LIU J，LIU J K，AN J J，et al.Precise crop classification based on multi‑features from time‑series Landsat 8 OLI

images and random forest algorithm［J］.Agricultural Research in the Arid Areas，2020，38（3）：281‑288.

［10］张东彦，戴震，徐新刚，等.基于时序Sentinel-2影像的现代农业园区作物分类研究［J］.红外与激光工程，2021，50（5）：262‑272.

ZHANG D Y，DAI Z，XU X G，et al.Crop classification of modern agricultural park based on time‑series Sentinel‑2 images［J］.Infrared and Laser Engineering，2021，50（5）：262‑272.

［11］申文明，王文杰，罗海江，等.基于决策树分类技术的遥感影像分类方法研究［J］.遥感技术与应用，2007，22（3）：333‑338.

SHEN W M，WANG W J，LUO H J，et al.Classification methods of remote sensing image based on decision tree technologies［J］.Remote Sensing Technology and Application，2007，22（3）：333‑338.

［12］田静，王卷乐，李一凡，等.基于决策树方法的蒙古高原土地覆盖遥感分类：以蒙古国中央省为例［J］.地球信息科学学报，2014，16（3）：460‑469.

TIAN J，WANG J L，LI Y F，et al.Land cover classification in Mongolian Plateau based on decision tree method：A case study in Tov Province，Mongolia［J］.Journal of Geo‑Information Science，2014，16（3）：460‑469.

［13］何昭欣，张淼，吴炳方，等.Google Earth Engine支持下的江苏省夏收作物遥感提取［J］.地球信息科学学报，2019，21（5）：752‑766.

HE Z X，ZHANG M，WU B F，et al.Extraction of summer crop in Jiangsu based on Google Earth Engine［J］.Journal of Geo‑Information Science，2019，21（5）：752‑766.

［14］黄健熙，侯矞焯，苏伟，等.基于GF-1 WFV数据的玉米与大豆种植面积提取方法［J］.农业工程学报，2017，33（7）：164‑170.

HUANG J X，HOU Y Z，SU W，et al.Mapping corn and soybean cropped area with GF‑1 WFV data［J］.Transactions of the Chinese Society of Agricultural Engineering，2017，33（7）：164‑170.

［15］王振兴，刘东，王敏.基于GEE平台和多维特征优选的粮食作物提取：以西辽河流域为例［J］.江苏农业科学，2023，51（21）：200‑208.

WANG Z X，LIU D，WANG M.Extraction of grain crop based on multidimensional feature optimization and GEE platform：Taking western Liaohe River Basin as an example［J］.Jiangsu Agricultural Sciences，2023，51（21）：200‑208. ［16］MATHER P M，KOCH M.Computer processing of remotely‑sensed images［M］.Hoboken：John Wiley &Sons，2010.

［17］CALDWELL A. Introduction to remote sensing［J］.ThePhotogrammetric Record，2003，18（103）：259.

［18］朱秀芳，潘耀忠，张锦水，等.训练样本对TM尺度小麦种植面积测量精度影响研究（Ⅰ）：训练样本与分类方法间分类精度响应关系研究［J］.遥感学报，2007，11 （6）：826‑837.

ZHU X F，PAN Y Z，ZHANG J S，et al.The effects of training samples on the wheat planting area measure

accuracy in TM scale（Ⅰ）：The accuracy response of different classifiers to training samples［J］.Journal of Remote Sensing，2007，11（6）：826‑837.

［19］阎静，王汶，李湘阁.利用神经网络方法提取水稻种植面积：以湖北省双季早稻为例［J］.遥感学报，2001，5 （3）：227‑230.

YAN J，WANG W，LI X G.Extracting the rice planting areas using an artificial neural network［J］.Journal of Remote Sensing，2001，5（3）：227‑230.

［20］BREIMAN L. Random forests［J］.Machine Learning，2001，45（1）：5‑32.

［21］王利民，刘佳，杨玲波，等.随机森林方法在玉米-大豆精细识别中的应用［J］.作物学报，2018，44（4）：569‑580.

WANG L M，LIU J，YANG L B，et al.Application of random forest method in maize‑soybean accurate identification［J］.Acta Agronomica Sinica，2018，44 （4）：569‑580.

［22］MCNAIRN H，PROTZ R.Mapping corn residue cover on agricultural fields in Oxford County,Ontario,using thematic mapper［J］.Canadian Journal of Remote Sensing，1993，19（2）：152‑159.

［23］QI J，HUETE A R，MORAN M S，et al.Interpretation of vegetation indices derived from multi‑temporal SPOT images［J］.Remote Sensing of Environment，1993，44（1）：89‑101.

［24］王娜.江苏省主要农作物遥感识别特征的时空效应研究［D］.北京：中国科学院大学（中国科学院遥感与数字地球研究所），2017.

WANG N.Temporal and spatial characteristics of crop identification features in Jiangsu Province［D］.Beijing：University of Chinese Academy of Sciences（Institute of Remote Sensing and Digital Earth Chinese Academy of Sciences），2017.

［25］金晶.基于纹理特征的高分辨率遥感影像分类方法研究［D］.长沙：中南大学，2013.

JIN J.Research of high resolution remote sensing image classification based on texture［D］.Changsha：Central South University，2013.

［26］韦春桃，王宁，张利恒，等.基于纹理特征的高分辨率遥感影像分类方法［J］.桂林理工大学学报，2013，33 （1）：80‑85.

WEI C T，WANG N，ZHANG L H，et al.Remote sensing image classification based on texture features［J］.Journal of Guilin University of Technology，2013,33（1）：80‑85.

［27］吴健平，杨星卫.遥感数据监督分类中训练样本的纯化［J］.国土资源遥感，1996，8（1）：36‑41.

WU J P，YANG X W.Purification of training samples in supervised classification of remote sensing data［J］.Remote Sensing for Natural Resources，1996，8（1）：36‑41.

［28］钟琪，罗津，齐述华.随机森林分类算法提取柑橘果园的样本数量敏感性分析［J］.江西科学，2019，37（5）：664‑669.

ZHONG Q，LUO J，QI S H.Sample sensitivity analysis of citrus orchards extracted by random forest classification algorithm［J］.Jiangxi Science，2019，37 （5）：664‑669.

［29］潘洪涛，王轩，王晓飞.训练样本对农作物遥感分类的精度影响研究［J］.红外与激光工程，2017，46（S1）：149‑156.

PAN H T，WANG X，WANG X F.Study on the effect of training samples on the accuracy of crop remote sensing classification［J］.Infrared and Laser Engineering，2017，46（S1）：149‑156.

［30］SHUKLA G，GARG R D，SRIVASTAVA H S，et al.Performance analysis of different predictive models for crop classification across an aridic to ustic area of Indian states［J］.Geocarto International，2018，33（3）：240‑259.

［31］张丰，熊桢，寇宁.高光谱遥感数据用于水稻精细分类研究［J］.武汉理工大学学报，2002，24（10）：36‑39.

ZHANG F，XIONG Z，KOU N.Airborne hyperspectral remote sensing image data is used for rice precise classification［J］.Journal of Wuhan University of Technology，2002，24（10）：36‑39.

［32］贾坤，李强子.农作物遥感分类特征变量选择研究现状与展望［J］.资源科学，2013，35（12）：2507‑2516.

JIA K，LI Q Z.Review of features selection in crop classification using remote sensing data［J］.Resources Science，2013，35（12）：2507‑2516.

［33］刘亮，姜小光，李显彬，等.利用高光谱遥感数据进行农作物分类方法研究［J］.中国科学院研究生院学报，2006，23（4）：484‑488.

LIU L，JIANG X G，LI X B，et al.Study on classification of agricultural crop by hyperspectral remote sensing data［J］.Journal of the Graduate School of the Chinese Academy of Sciences，2006，23（4）：484‑488.

［34］王文静，张霞，赵银娣，等.综合多特征的Landsat 8时序遥感图像棉花分类方法［J］.遥感学报，2017，21 （1）：115‑124.

WANG W J，ZHANG X，ZHAO Y D，et al.Cotton extraction method of integrated multi‑features based on multitemporal Landsat 8 images［J］.Journal of Remote Sensing，2017，21（1）：115‑124.

[1]	ZHANG Chaoran, PENG Yi, YANG Kaili, YUAN Ningge, LI Yuanjin, LIU Xiaojuan, WU Huili. Estimation of Rice Panicle Biomass by Remotely Sensed Photosynthetically Active Absorption [J]. Journal of Henan Agricultural Sciences, 2024, 53(6): 154-161.
[2]	LI Min, GUO Leifeng, WANG Ruili. Application of Unmanned Aerial Vehicle（UAV）Remote Sensing in the Study of Potato Field Phenotype [J]. Journal of Henan Agricultural Sciences, 2023, 52(9): 24-32.
[3]	ZHANG Keqian, CHENG Gang, WU Wei, SONG Xiangyang, ZHANG Ziqian, YAO Shun, WU Shuai. Extraction of Winter Wheat Planting Area Based on Fused Active and Passive Remote Sensing Images [J]. Journal of Henan Agricultural Sciences, 2023, 52(6): 160-171.
[4]	ZANG Hecang, WANG Congsheng, ZHAO Qiaoli, ZHAO Qing, ZHANG Jie, LI Guoqiang, ZHENG Guoqing. Study on Automatic Classification Method of Wheat Lodging Based on Deep Learning [J]. Journal of Henan Agricultural Sciences, 2023, 52(11): 167-173.
[5]	ZHENG Zirui, ZHAO Huijie, WEI Panpan, FANG Peng, WANG Laigang, XU Shaobo. Integration of Multi‑Source Remote Sensing Data and Temporal Spectral Features of Growth Stages for Rice Planting Area Extraction [J]. Journal of Henan Agricultural Sciences, 2023, 52(10): 153-161.
[6]	XU Xibo, LÜ Minghui, WANG Haihui, ZHOU Zhongke, PENG Yuanxin, YAN Xuewen. Estimation of Soil Salt Content of Cultivated Land during Bare Soil Period Using Spatial Random Forest Model and Remote Sensing Images [J]. Journal of Henan Agricultural Sciences, 2022, 51(5): 162-170.
[7]	WANG Laigang, XU Shaobo, LI Shimin, GUO Yan, CHENG Yongzheng, HE Jia. Study on Automatic Extraction Algorithm of Farmland Flood Disaster Information Based on Remote Sensing [J]. Journal of Henan Agricultural Sciences, 2022, 51(11): 163-170.
[8]	MO Jiacai, PENG Yi, LIU Xiaojuan, WANG Jing. Estimation of Heading Time for Rice of Different Cultivars Based on UAV Remote Sensing Images [J]. Journal of Henan Agricultural Sciences, 2022, 51(1): 154-161.
[9]	WEI Qingwei, ZHU Liming , WANG Fuzhou. A Method for Monitoring the Critical Growth Stages of Winter Wheat by Using Near‑earth Remote Sensing [J]. Journal of Henan Agricultural Sciences, 2021, 50(10): 163-171.
[10]	WANG Yihao, WU Yongfeng, ZHANG Liting, ZHANG Jinshui, LI Chongrui, ZHANG Xiaoxu. Drought Monitoring of Spring Maize in Northeast China Based on TVDI———Taking the 2018 for Instance [J]. Journal of Henan Agricultural Sciences, 2020, 49(3): 167-180.