Crop classification from Sentinel-2-derived vegetation indices using ensemble learning

Sonobe, Rei; Yamaya, Yuki; Tani, Hiroshi; Wang, Xiufeng; Kobayashi, Nobuyuki; Mochizuki, Kiyofumi

doi:10.1117/1.jrs.12.026019

Cited by 141 publications

(79 citation statements)

References 79 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Sentinel-2 carries a MultiSpectral Instrument (MSI) with 13 bands in the visible, near infrared, and shortwave infrared part of the spectrum. These state-of-the-art specifications have been used in recent research on crop classification [16][17][18].…”

Section: Introductionmentioning

confidence: 99%

Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium

et al. 2018

View full text Add to dashboard Cite

A timely inventory of agricultural areas and crop types is an essential requirement for ensuring global food security and allowing early crop monitoring practices. Satellite remote sensing has proven to be an increasingly more reliable tool to identify crop types. With the Copernicus program and its Sentinel satellites, a growing source of satellite remote sensing data is publicly available at no charge. Here, we used joint Sentinel-1 radar and Sentinel-2 optical imagery to create a crop map for Belgium. To ensure homogenous radar and optical inputs across the country, Sentinel-1 12-day backscatter mosaics were created after incidence angle normalization, and Sentinel-2 normalized difference vegetation index (NDVI) images were smoothed to yield 10-daily cloud-free mosaics. An optimized random forest classifier predicted the eight crop types with a maximum accuracy of 82% and a kappa coefficient of 0.77. We found that a combination of radar and optical imagery always outperformed a classification based on single-sensor inputs, and that classification performance increased throughout the season until July, when differences between crop types were largest. Furthermore, we showed that the concept of classification confidence derived from the random forest classifier provided insight into the reliability of the predicted class for each pixel, clearly showing that parcel borders have a lower classification confidence. We concluded that the synergistic use of radar and optical data for crop classification led to richer information increasing classification accuracies compared to optical-only classification. Further work should focus on object-level classification and crop monitoring to exploit the rich potential of combined radar and optical observations.

show abstract

Section: Introductionmentioning

confidence: 99%

Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium

et al. 2018

View full text Add to dashboard Cite

show abstract

“…For instance, Qiu et al[2] mapped winter wheat in North China using MODIS data, and reported a producer accuracy of 82.77% and a user accuracy of 81.49% when compared to the classification results from Landsat images. Landsat and Sentinel images are data sources which are often used in agricultural remote sensing [25][26][27][28]. However, Landsat or Sentinel images have generally been used at smaller scales in previous research, because of the difficulties in image pre-processing [29,30].…”

mentioning

confidence: 99%

Mapping Winter Crops in China with Multi-Source Satellite Imagery and Phenology-Based Algorithm

et al. 2019

View full text Add to dashboard Cite

Timely and accurate mapping of winter crop planting areas in China is important for food security assessment at a national level. Time-series of vegetation indices, such as the normalized difference vegetation index (NDVI), are widely used for crop mapping, as they can characterize the growth cycle of crops. However, with the moderate spatial resolution optical imagery acquired by Landsat and Sentinel-2, it is difficult to obtain complete time-series curves for vegetation indices due to the influence of the revisit cycle of the satellite and weather conditions. Therefore, in this study, we propose a method for compositing the multi-temporal NDVI, in order to map winter crop planting areas with the Landsat-7 and -8 and Sentinel-2 optical images. The algorithm composites the multi-temporal NDVI into three key values, according to two time-windows-a period of low NDVI values and a period of high NDVI values-for the winter crops. First, we identify the two time-windows, according to the time-series of the NDVI obtained from daily Moderate Resolution Imaging Spectroradiometer observations. Second, the 30 m spatial resolution multi-temporal NDVI curve, derived from the Landsat-7 and -8 and Sentinel-2 optical images, is composited by selecting the maximal value in the high NDVI value period, and the minimal and median values in the low NDVI value period, using an algorithm of the Google Earth Engine. Third, a decision tree classification method is utilized to perform the winter crop classification at a pixel level. The results indicate that this method is effective for the large-scale mapping of winter crops. In the study area, the area of winter crops in 2018 was determined to be 207,641 km 2 , with an overall accuracy of 96.22% and a kappa coefficient of 0.93. The method proposed in this paper is expected to contribute to the rapid and accurate mapping of winter crops in large-scale applications and analyses.Resolution Imaging Spectroradiometer (MODIS) Land Cover Collections [12], and the 1 km Global Land Cover 2000 (GLC2000) [13]). However, more specific information on winter crops is still limited in existing land-cover products [2,14]. On the other hand, the mapping of winter crop planting areas using the traditional agricultural census approach requires a good deal of manpower, money, and time [15]. Further, the timeliness of the traditional approach is poor. For example, the Chinese government generally publishes this information at least six months after the crops have been harvested. Furthermore, the quality of the survey data tends to suffer from the influence of human errors [3].Remote sensing techniques provide a very effective means to map crops [2,11,16,17], due to their fast responses, periodic observations, wide field of view, and low cost [18][19][20]. However, there are still some challenges associated with large-scale (e.g., provincial-or national-scale) winter crop mapping by remote sensing. MODIS data was the major data source for large-scale crop mapping in previous studies, due to its high revisit...

show abstract

“…On invaded landscapes, Parthenium weed expands more rapidly than native plants [60]. Spectral bands alone are not enough to achieve reliable mapping accuracies [61]. Increasing data dimensionality by combining, among others, Sentinel-2 image bands, vegetation indices and other variable types and applying an appropriate feature selection approach, higher Pathenium mapping accuracy can be achieved.…”

Section: Implications Of Findings In Parthenium Weed Managementmentioning

confidence: 99%

Feature Selection on Sentinel-2 Multispectral Imagery for Mapping a Landscape Infested by Parthenium Weed

et al. 2019

View full text Add to dashboard Cite

In the recent past, the volume of spatial datasets has significantly increased. This is attributed to, among other factors, higher sensor temporal resolutions of the recently launched satellites. The increased data, combined with the computation and possible derivation of a large number of indices, may lead to high multi-collinearity and redundant features that compromise the performance of classifiers. Using dimension reduction algorithms, a subset of these features can be selected, hence increasing their predictive potential. In this regard, an investigation into the application of feature selection techniques on multi-temporal multispectral datasets such as Sentinel-2 is valuable in vegetation mapping. In this study, ten feature selection methods belonging to five groups (Similarity-based, statistical-based, Sparse learning based, Information theoretical based, and wrappers methods) were compared based on f-score and data size for mapping a landscape infested by the Parthenium weed (Parthenium hysterophorus). Overall, results showed that ReliefF (a Similarity-based approach) was the best performing feature selection method as demonstrated by the high f-score values of Parthenium weed and a small size of optimal features selected. Although svm-b (a wrapper method) yielded the highest accuracies, the size of optimal subset of selected features was quite large. Results also showed that data size affects the performance of feature selection algorithms, except for statistically-based methods such as Gini-index and F-score and svm-b. Findings in this study provide a guidance on the application of feature selection methods for accurate mapping of invasive plant species in general and Parthenium weed, in particular, using new multispectral imagery with high temporal resolution.

show abstract

Crop classification from Sentinel-2-derived vegetation indices using ensemble learning

Cited by 141 publications

References 79 publications

Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium

Synergistic Use of Radar Sentinel-1 and Optical Sentinel-2 Imagery for Crop Mapping: A Case Study for Belgium

Mapping Winter Crops in China with Multi-Source Satellite Imagery and Phenology-Based Algorithm

Feature Selection on Sentinel-2 Multispectral Imagery for Mapping a Landscape Infested by Parthenium Weed

Contact Info

Product

Resources

About