Self-Learning Based Land-Cover Classification Using Sequential Class Patterns from Past Land-Cover Maps

被引:21
作者
Kim, Yeseul [1 ]
Park, No-Wook [1 ]
Lee, Kyung-Do [2 ]
机构
[1] Inha Univ, Dept Geoinformat Engn, Incheon 22212, South Korea
[2] Rural Dev Adm, Natl Inst Agr Sci, Wonju 55365, South Korea
关键词
classification; self-learning; training data; crop; RANDOM FOREST CLASSIFICATION; SUPPORT VECTOR MACHINE; REMOTE-SENSING IMAGES; TIME-SERIES; CROP CLASSIFICATION; NDVI DATA; MODIS; ALGORITHMS; ROTATIONS; ACCURACY;
D O I
10.3390/rs9090921
中图分类号
X [环境科学、安全科学];
学科分类号
08 ; 0830 ;
摘要
To improve the accuracy of classification with a small amount of training data, this paper presents a self-learning approach that defines class labels from sequential patterns using a series of past land-cover maps. By stacking past land-cover maps, unique sequence rule information from sequential change patterns of land-covers is first generated, and a rule-based class label image is then prepared for a given time. After the most informative pixels with high uncertainty are selected from the initial classification, rule-based class labels are assigned to the selected pixels. These newly labeled pixels are added to training data, which then undergo an iterative classification process until a stopping criterion is reached. Time-series MODIS NDVI data sets and cropland data layers (CDLs) from the past five years are used for the classification of various crop types in Kansas. From the experiment results, it is found that once the rule-based labels are derived from past CDLs, the labeled informative pixels could be properly defined without analyst intervention. Regardless of different combinations of past CDLs, adding these labeled informative pixels to training data increased classification accuracy and the maximum improvement of 8.34 percentage points in overall accuracy was achieved when using three CDLs, compared to the initial classification result using a small amount of training data. Using more than three consecutive CDLs showed slightly better classification accuracy than when using two CDLs (minimum and maximum increases were 1.56 and 2.82 percentage points, respectively). From a practical viewpoint, using three or four CDLs was the best choice for this study area. Based on these experiment results, the presented approach could be applied effectively to areas with insufficient training data but access to past land-cover maps. However, further consideration should be given to select the optimal number of past land-cover maps and reduce the impact of errors of rule-based labels.
引用
收藏
页数:20
相关论文
共 48 条
[1]   Microbial communities and enzyme activities in soils under alternative crop rotations compared to wheat-fallow for the Central Great Plains [J].
Acosta-Martinez, Veronica ;
Mikha, Maysoon M. ;
Vigil, Merle F. .
APPLIED SOIL ECOLOGY, 2007, 37 (1-2) :41-52
[2]  
[Anonymous], 2005, SEMISUPERVISED LEARN
[3]  
[Anonymous], 2006, BOOK REV IEEE T NEUR
[4]  
[Anonymous], 2010, 1648 U WISC MAD
[5]   Classifying Complex Mountainous Forests with L-Band SAR and Landsat Data Integration: A Comparison among Different Machine Learning Methods in the Hyrcanian Forest [J].
Attarchi, Sara ;
Gloaguen, Richard .
REMOTE SENSING, 2014, 6 (05) :3624-3647
[6]  
Blum A., 1998, Proceedings of the Eleventh Annual Conference on Computational Learning Theory, P92, DOI 10.1145/279943.279962
[7]   Monitoring US agriculture: the US Department of Agriculture, National Agricultural Statistics Service, Cropland Data Layer Program [J].
Boryan, Claire ;
Yang, Zhengwei ;
Mueller, Rick ;
Craig, Mike .
GEOCARTO INTERNATIONAL, 2011, 26 (05) :341-358
[8]   A Novel Context-Sensitive Semisupervised SVM Classifier Robust to Mislabeled Training Samples [J].
Bruzzone, Lorenzo ;
Persello, Claudio .
IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, 2009, 47 (07) :2142-2154
[9]   A simple method for reconstructing a high-quality NDVI time-series data set based on the Savitzky-Golay filter [J].
Chen, J ;
Jönsson, P ;
Tamura, M ;
Gu, ZH ;
Matsushita, B ;
Eklundh, L .
REMOTE SENSING OF ENVIRONMENT, 2004, 91 (3-4) :332-344
[10]   Temporal segmentation of MODIS time series for improving crop classification in Central Asian irrigation systems [J].
Conrad, Christopher ;
Colditz, Rene R. ;
Dech, Stefan ;
Klein, Doris ;
Vlek, Paul L. G. .
INTERNATIONAL JOURNAL OF REMOTE SENSING, 2011, 32 (23) :8763-8778