Mapping Rice Paddy Based on Machine Learning with Sentinel-2 Multi-Temporal Data: Model Comparison and Transferability

Rice is an important agricultural crop in the Southwest Hilly Area, China, but there has been a lack of efficient and accurate monitoring methods in the region. Recently, convolutional neural networks (CNNs) have obtained considerable achievements in the remote sensing community. However, it has not been widely used in mapping a rice paddy, and most studies lack the comparison of classification effectiveness and efficiency between CNNs and other classic machine learning models and their transferability. This study aims to develop various machine learning classification models with remote sensing data for comparing the local accuracy of classifiers and evaluating the transferability of pretrained classifiers. Therefore, two types of experiments were designed: local classification experiments and model transferability experiments. These experiments were conducted using cloud-free Sentinel-2 multi-temporal data in Banan District and Zhongxian County, typical hilly areas of Southwestern China. A pure pixel extraction algorithm was designed based on land-use vector data and a Google Earth Online image. Four convolutional neural network (CNN) algorithms (one-dimensional (Conv-1D), two-dimensional (Conv-2D) and three-dimensional (Conv-3D_1 and Conv-3D_2) convolutional neural networks) were developed and compared with four widely used classifiers (random forest (RF), extreme gradient boosting (XGBoost), support vector machine (SVM) and multilayer perceptron (MLP)). Recall, precision, overall accuracy (OA) and F1 score were applied to evaluate classification accuracy. The results showed that Conv-2D performed best in local classification experiments with OA of 93.14% and F1 score of 0.8552 in Banan District, OA of 92.53% and F1 score of 0.8399 in Zhongxian County. CNN-based models except Conv-1D provided more desirable performance than non-CNN classifiers. Besides, among the non-CNN classifiers, XGBoost received the best result with OA of 89.73% and F1 score of 0.7742 in Banan District, SVM received the best result with OA of 88.57% and F1 score of 0.7538 in Zhongxian County. In model transferability experiments, almost all CNN classifiers had low transferability. RF and XGBoost models have achieved acceptable F1 scores for transfer (RF = 0.6673 and 0.6469, XGBoost = 0.7171 and 0.6709, respectively).

[1]  Dailiang Peng,et al.  Detection and estimation of mixed paddy rice cropping patterns with MODIS data , 2011, Int. J. Appl. Earth Obs. Geoinformation.

[2]  Peter M. Atkinson,et al.  Mapping paddy rice fields by applying machine learning algorithms to multi-temporal Sentinel-1A and Landsat data , 2018 .

[3]  Yi-Shiang Shiu,et al.  Mapping paddy rice agriculture in a highly fragmented area using a geographic information system object-based post classification process , 2012 .

[4]  A. Huete,et al.  A Modified Soil Adjusted Vegetation Index , 1994 .

[5]  William J. Emery,et al.  Contextually guided very-high-resolution imagery classification with semantic segments , 2017 .

[6]  Derek T. Anderson,et al.  Comprehensive survey of deep learning in remote sensing: theories, tools, and challenges for the community , 2017 .

[7]  Jinwei Dong,et al.  Mapping paddy rice planting area in cold temperate climate region through analysis of time series Landsat 8 (OLI), Landsat 7 (ETM+) and MODIS imagery. , 2015, ISPRS journal of photogrammetry and remote sensing : official publication of the International Society for Photogrammetry and Remote Sensing.

[8]  Julio Lumbreras,et al.  A comprehensive approach for the evaluation and comparison of emission inventories in Madrid , 2016 .

[9]  Jie Chang,et al.  Methane emissions from rice paddies natural wetlands, lakes in China: synthesis new estimate , 2013, Global change biology.

[10]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[11]  Rajendra Prasad,et al.  Knowledge-based decision tree approach for mapping spatial distribution of rice crop using C-band synthetic aperture radar-derived information , 2017 .

[12]  Tingting Fu,et al.  Ecosystem health assessment based on analysis of a land use database , 2013 .

[13]  S. Henneberry,et al.  Socio-cultural roots of rural settlement dispersion in Sichuan Basin: The perspective of Chinese lineage , 2019, Land Use Policy.

[14]  Jinwei Dong,et al.  Mapping paddy rice distribution using multi-temporal Landsat imagery in the Sanjiang Plain, northeast China , 2016, Frontiers of Earth Science.

[15]  Xiaodong Li,et al.  Water Bodies' Mapping from Sentinel-2 Imagery with Modified Normalized Difference Water Index at 10-m Spatial Resolution Produced by Sharpening the SWIR Band , 2016, Remote. Sens..

[16]  Yaozhong Pan,et al.  A highly efficient temporal-spatial probability synthesized model from multi-temporal remote sensing for paddy rice identification , 2017 .

[17]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[18]  Bo Du,et al.  Deep Learning for Remote Sensing Data: A Technical Tutorial on the State of the Art , 2016, IEEE Geoscience and Remote Sensing Magazine.

[19]  Zhanqing Li,et al.  Estimating 1-km-resolution PM2.5 concentrations across China using the space-time random forest approach , 2019, Remote Sensing of Environment.

[20]  B. Bouman,et al.  Rice and Water , 2007 .

[21]  Changsheng Li,et al.  Observation of flooding and rice transplanting of paddy rice fields at the site to landscape scales in China using VEGETATION sensor data , 2002 .

[22]  T. Motohka,et al.  Evaluation of optical satellite remote sensing for rice paddy phenology in monsoon Asia using a continuous in situ dataset , 2009 .

[23]  Hang Zhou,et al.  Deep learning based multi-temporal crop classification , 2019, Remote Sensing of Environment.

[24]  P. Swain,et al.  Neural Network Approaches Versus Statistical Methods In Classification Of Multisource Remote Sensing Data , 1990 .

[25]  Giorgos Mountrakis,et al.  A meta-analysis of remote sensing research on supervised pixel-based land-cover image classification processes: General guidelines for practitioners and future research , 2016 .

[26]  Changsheng Li,et al.  Mapping paddy rice agriculture in South and Southeast Asia using multi-temporal MODIS images , 2006 .

[27]  Sebastian Varela,et al.  Forecasting maize yield at field scale based on high-resolution satellite imagery , 2018, Biosystems Engineering.

[28]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[29]  Abdulhakim M. Abdi,et al.  Land cover and land use classification performance of machine learning algorithms in a boreal landscape using Sentinel-2 data , 2019, GIScience & Remote Sensing.

[30]  G. Foody,et al.  Measuring River Wetted Width From Remotely Sensed Imagery at the Subpixel Scale With a Deep Convolutional Neural Network , 2019, Water Resources Research.

[31]  Rameen Abdal,et al.  UFCN: a fully convolutional neural network for road extraction in RGB imagery acquired by remote sensing from an unmanned aerial vehicle , 2018 .

[32]  Zhanqing Li,et al.  Satellite-derived 1-km-resolution PM1 concentrations from 2014 to 2018 across China. , 2019, Environmental science & technology.

[33]  Hiroyuki Ohno,et al.  Spatio-temporal distribution of rice phenology and cropping systems in the Mekong Delta with special reference to the seasonal water flow of the Mekong and Bassac rivers , 2006 .

[34]  Nataliia Kussul,et al.  Deep Learning Classification of Land Cover and Crop Types Using Remote Sensing Data , 2017, IEEE Geoscience and Remote Sensing Letters.

[35]  Xiuping Jia,et al.  Deep Feature Extraction and Classification of Hyperspectral Images Based on Convolutional Neural Networks , 2016, IEEE Transactions on Geoscience and Remote Sensing.

[36]  Matthias Drusch,et al.  Sentinel-2: ESA's Optical High-Resolution Mission for GMES Operational Services , 2012 .

[37]  Samsuzana Abd Aziz,et al.  Quantifying sub-pixel signature of paddy rice field using an artificial neural network , 2009 .

[38]  Florent Lafarge,et al.  Pyramid scene parsing network in 3D: Improving semantic segmentation of point clouds with multi-scale contextual information , 2019, ISPRS Journal of Photogrammetry and Remote Sensing.

[39]  Hui Lin,et al.  Mapping Paddy Rice Using a Convolutional Neural Network (CNN) with Landsat 8 Datasets in the Dongting Lake Area, China , 2018, Remote. Sens..

[40]  Shanjun Mao,et al.  Spectral–spatial classification of hyperspectral images using deep convolutional neural networks , 2015 .

[41]  Yang Chen,et al.  Extraction of Urban Water Bodies from High-Resolution Remote-Sensing Imagery Using Deep Learning , 2018 .

[42]  Matthew L. Clark,et al.  One-Dimensional Convolutional Neural Network Land-Cover Classification of Multi-Seasonal Hyperspectral Imagery in the San Francisco Bay Area, California , 2017, Remote. Sens..

[43]  Weixing Cao,et al.  Common Spectral Bands and Optimum Vegetation Indices for Monitoring Leaf Nitrogen Accumulation in Rice and Wheat , 2012 .

[44]  Hanqiu Xu.  Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery , 2006 .

[45]  Fan Zhang,et al.  Deep Convolutional Neural Networks for Hyperspectral Image Classification , 2015, J. Sensors.

[46]  Lin Sun,et al.  Improved 1 km resolution PM2.5 estimates across China using enhanced space–time extremely randomized trees , 2020 .

[47]  Meng Lu,et al.  A scale robust convolutional neural network for automatic building extraction from aerial and satellite imagery , 2018, International Journal of Remote Sensing.

[48]  Mario Chica-Olmo,et al.  An assessment of the effectiveness of a random forest classifier for land-cover classification , 2012 .

[49]  Yi Li,et al.  Winter wheat mapping using a random forest classifier combined with multi-temporal and multi-sensor data , 2018, Int. J. Digit. Earth.

[50]  Vasant P. Gandhi,et al.  Food demand and the food security challenge with rapid economic growth in the emerging economies of India and China , 2014 .

[51]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[52]  Hankui K. Zhang,et al.  Finer resolution observation and monitoring of global land cover: first mapping results with Landsat TM and ETM+ data , 2013 .

[53]  Clement Atzberger,et al.  First Experience with Sentinel-2 Data for Crop and Tree Species Classifications in Central Europe , 2016, Remote. Sens..

[54]  Yong-sheng Wang,et al.  The Challenges and Strategies of Food Security under Rapid Urbanization in China , 2019, Sustainability.

[55]  Changsheng Li,et al.  Mapping paddy rice agriculture in southern China using multi-temporal MODIS images , 2005 .

[56]  S. Robinson,et al.  Food Security: The Challenge of Feeding 9 Billion People , 2010, Science.

[57]  Jinwei Dong,et al.  Mapping paddy rice planting areas through time series analysis of MODIS land surface temperature and vegetation index data. , 2015, ISPRS journal of photogrammetry and remote sensing : official publication of the International Society for Photogrammetry and Remote Sensing.