基于时序光谱重构的卷积神经网络遥感农作物分类

doi:10.7523/j.issn.2095-6134.2020.05.006

摘要/Abstract

摘要： 当前，基于时序特征提取的农作物遥感分类方法需要较多先验知识及人工干预，难以自动化，且易因忽略部分有效特征而导致精度降低。针对这些问题，提出基于时序光谱重构的卷积神经网络农作物分类法。为充分利用时间序列多光谱中丰富的作物物候与多光谱信息，对每个地面像元构造以时间维为纵轴、光谱维为横轴的时序光谱图，采用Adam梯度下降法与Dropout 40%连接率优化后的卷积神经网络对时序光谱图进行分类。对比实验结果表明，该方法可有效减少“椒盐”噪声的产生，且地块边界轮廓线清晰，总体分类精度达到95.12%，高于时间序列多光谱+随机森林（88.58%）、时间序列NDVI+随机森林（90.25%）、时间序列NDVI+卷积神经网络（91.79%）等对照实验组；对于“异物同谱”现象明显的春玉米与番茄，该方法的F₁-score分别达到95.9%与89.9%，相比各对照组均有较大幅度的提高。该研究结果可为遥感农作物的自动化精细制图提供参考。

关键词: 农作物分类, 卷积神经网络, 遥感, Sentinel-2A, 时间序列, 特征提取

Abstract: At the present, the crop classification method based on time series feature extraction needs much prior knowledge and manual intervention. It is difficult to automate the classification and it is easy to reduce the accuracy due to neglecting some effective features. We propose a convolutional neural network crop classification method based on time series spectral reconstruction. The method constructs a time-series image for every ground pixel, with the time dimension as the vertical axis and the spectral dimension as the horizontal axis, and then uses CNN, optimized by using the Adam gradient descent method and 40% connection rate dropout, to classify the time-series images. The experimental results show that the method reduces the salt-and-pepper noise, and the boundary of the plot is clear. The overall classification accuracy reaches 95.12%, which is higher than those of the time series multispectral + random forest method(88.58%), time-series NDVI + random forest method(90.25%), and time-series NDVI + convolutional neural network method(91.79%). For spring corn and tomato with close spectral similarity, the F₁-scores of our method reach 95.9% and 89.9% respectively, which are significantly improved compared with the control groups. The results of this study provide useful information for the automatic and fine mapping of croplands.

Key words: crop classification, convolutional neural network, remote sensing, Sentinel-2A, time series, feature extraction

中图分类号:

TP79

冯齐心, 杨辽, 王伟胜, 陈桃, 黄双燕. 基于时序光谱重构的卷积神经网络遥感农作物分类[J]. 中国科学院大学学报, 2020, 37(5): 619-628.

FENG Qixin, YANG Liao, WANG Weisheng, CHEN Tao, HUANG Shuangyan. CNN remote sensing crop classification based on time series spectral reconstruction[J]. , 2020, 37(5): 619-628.

参考文献

[1] Say S M, Keskin M, Sehri M, et al. Adoption of precision agriculture technologies in developed and developing countries[C]//International Science and Technology Conferecce. Berlin, 2017:41-49.
[2] 方向明, 李姣媛. 精准农业:发展效益、国际经验与中国实践[J]. 农业经济问题, 2018(11):28-37.
[3] Bargiel D. A new method for crop classification combining time series of radar images and crop phenology information[J]. Remote Sensing of Environment, 2017, 198:369-383.
[4] 张佳华, 胡小夏, 刘学锋, 等. 基于MODIS数据提取华北典型区冬小麦种植面积[J]. 中国科学院研究生院学报, 2013, 30(5):637-643.
[5] 史舟, 梁宗正, 杨媛媛, 等. 农业遥感研究现状与展望[J]. 农业机械学报, 2015, 46(2):247-260.
[6] 史飞飞, 高小红, 杨灵玉, 等. 基于HJ-1A高光谱遥感数据的湟水流域典型农作物分类研究[J]. 遥感技术与应用, 2017, 32(2):206-217.
[7] 王文静, 张霞, 赵银娣, 等. 综合多特征的Landsat8时序遥感图像棉花分类方法[J]. 遥感学报, 2017, 21(1):115-124.
[8] 史飞飞, 雷春苗, 肖建设, 等. 基于多源遥感数据的复杂地形区农作物分类[J]. 地理与地理信息科学, 2018, 34(5):49-55,2.
[9] Cai Y, Guan K, Peng J, et al. A high-performance and in-season classification system of field-level crop types using time-series Landsat data and a machine learning approach[J]. Remote Sensing of Environment, 2018, 210:35-47.
[10] 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.
[11] Chauhan H J, Mohan B K. Measures to improve crop classification using remotely sensed hyperion hyperspectral imagery[C]//2012 International Conference on Communications, Devices and Intelligent Systems (CODIS). IEEE, 2012:596-599.
[12] Senf C, Leitão P J, Pflugmacher D, et al. Mapping land cover in complex Mediterranean landscapes using Landsat:improved classification accuracies from integrating multi-seasonal and synthetic imagery[J]. Remote Sensing of Environment, 2015, 156:527-536.
[13] 翟涌光, 屈忠义. 基于非线性降维时序遥感影像的作物分类[J]. 农业工程学报, 2018, 34(19):177-183.
[14] Löw F, Michel U, Dech S, et al. Impact of feature selection on the accuracy and spatial uncertainty of per-field crop classification using support vector machines[J]. Isprs Journal of Photogrammetry & Remote Sensing, 2013, 85(6):102-119.
[15] Tewkesbury A P, Comber A J, Tate N J, et al. A critical synthesis of remotely sensed optical image change detection techniques[J]. Remote Sensing of Environment, 2015, 160:1-14.
[16] Ji S, Zhang C, Xu A, et al. 3D convolutional neural networks for crop classification with multi-temporal remote sensing images[J]. Remote Sensing, 2018, 10(2):75.
[17] 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.
[18] 郭昱杉, 刘庆生, 刘高焕, 等. 基于MODIS时序NDVI主要农作物种植信息提取研究[J]. 自然资源学报, 2017, 32(10):1808-1818.
[19] Belgiu M, Csillik O. Sentinel-2 cropland mapping using pixel-based and object-based time-weighted dynamic time warping analysis[J]. Remote Sensing of Environment, 2018, 204:509-523.
[20] Ma X, Fu A, Wang J, et al. Hyperspectral image classification based on deep deconvolution network with skip architecture[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(8):4781-4791.
[21] 闫苗, 赵红东, 李宇海, 等. 基于卷积神经网络的高光谱遥感地物多分类识别[J]. 激光与光电子学进展, 2019, 56(2):191-198.
[22] 江璐, 赵彤, 吴敏. 基于深度卷积神经网络的指纹纹型分类算法[J]. 中国科学院大学学报, 2016, 33(6):808-814.
[23] 吕俊奇, 邱卫根, 张立臣, 等. 多层卷积特征融合的行人检测[J]. 计算机工程与设计, 2018, 39(11):3481-3485.
[24] 欧攀, 张正, 路奎, 等. 基于卷积神经网络的遥感图像目标检测[J]. 激光与光电子学进展, 2019, 56(5):1-12.
[25] Krizhevsky A, Sutskever I, Hinton G E. ImageNet classification with deep convolutional neural networks[C]//Advances in Neural Information Processing Systems. Lake Tahoe, 2012:1097-1105.
[26] 陈超, 齐峰. 卷积神经网络的发展及其在计算机视觉领域中的应用综述[J]. 计算机科学, 2019(3):63-73.
[27] Xu X, Li W, Ran Q, et al. Multisource remote sensing data classification based on convolutional neural network[J]. IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(2):937-949.
[28] Adelabu S, Mutanga O, Adam E. Evaluating the impact of red-edge band from Rapideye image for classifying insect defoliation levels[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014, 95:34-41.
[29] 郑阳, 吴炳方, 张淼. Sentinel-2数据的冬小麦地上干生物量估算及评价[J]. 遥感学报, 2017, 21(2):318-328.
[30] Rapinel S, Mony C, Lecoq L, et al. Evaluation of Sentinel-2 time-series for mapping floodplain grassland plant communities[J]. Remote Sensing of Environment, 2019, 223:115-129.
[31] Truong T D, Nguyen V T, Tran M T. Lightweight deep convolutional network for tiny object recognition[C]//International Conference on Pattern Recognition Applications and Methods. Funchal, 2018:675-682.
[32] Hinton G E, Srivastava N, Krizhevsky A, et al. Improving neural networks by preventing co-adaptation of feature detectors[J], Comput Science, 2012, 3(4):212-223.
[33] Srivastava N, Hinton G E, Krizhevsky A, et al. Dropout:a simple way to prevent neural networks from overfitting[J]. Journal of Machine Learning Research, 2014, 15(1):1929-1958.
[34] Zhong L, Hu L, Zhou H. Deep learning based multi-temporal crop classifcation[J]. Remote Sensing of Environment, 2019, 221:430-443.
[35] Kussul N, Lavreniuk M, Skakun S, et al. Deep learning classification of land cover and crop types using remote sensing data[J]. IEEE Geoscience and Remote Sensing Letters, 2017, 14(5):778-782.
[36] 张兵. 遥感大数据时代与智能信息提取[J]. 武汉大学学报(信息科学版), 2018, 43(12):1861-1871.
[37] Tuia D, Marcos D, Camps-Valls G. Multi-temporal and multi-source remote sensing image classification by nonlinear relative normalization[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016, 120:1-12.