Crop maps are essential inputs for the agricultural planning done at various governmental and agribusinesses agencies. Remote sensing offers timely and costs efficient technologies to identify and map crop types over large areas. Among the plethora of classification methods, Support Vector Machine (SVM) and Random Forest (RF) are widely used because of their proven performance. In this work, we study the synergic use of both methods by introducing a random forest kernel (RFK) in an SVM classifier. A time series of multispectral WorldView-2 images acquired over Mali (West Africa) in 2014 was used to develop our case study. Ground truth containing five common crop classes (cotton, maize, millet, peanut, and sorghum) were collected at 45 farms and used to train and test the classifiers. An SVM with the standard Radial Basis Function (RBF) kernel, a RF, and an SVM-RFK were trained and tested over 10 random training and test subsets generated from the ground data. Results show that the newly proposed SVM-RFK classifier can compete with both RF and SVM-RBF. The overall accuracies based on the spectral bands only are of 83, 82 and 83% respectively. Adding vegetation indices to the analysis result in the classification accuracy of 82, 81 and 84% for SVM-RFK, RF, and SVM-RBF respectively. Overall, it can be observed that the newly tested RFK can compete with SVM-RBF and RF classifiers in terms of classification accuracy.
The Medium Resolution Imaging Spectrometer, MERIS, on board of ENVISAT-1 fulfils the information gap between the current high and low spatial resolution sensors. In this respect, the use of MERIS full resolution data (300 m pixel size) has a great potential for regional and global land cover mapping. However, the spectral and temporal resolutions of MERIS (15 narrow bands and a revisit time of 2-3 days, respectively) might be further exploited in order to get land cover information at a more detailed scale. The performance of MERIS for extracting sub-pixel land cover information was evaluated in this study. An iterative linear spectral unmixing method designed to optimize the number of endmembers per pixel was used to classify 2 MERIS full resolution images acquired over The Netherlands. The latest version of the Dutch land use database, the LGN5, was used as a reference dataset both for the validation and for the selection of the endmembers. This dataset was first thematically aggregated to the main 9 land cover types and then spatially aggregated from its original 25m to 300m. Because the fractions of the different land cover types present in each MERIS pixel were computed during the aggregation, a sub-pixel accuracy assessment could be done (in addition to the traditional assessment based on a hard classification). Results pointed out that MERIS has a great potential for providing sub-pixel land cover information because the classification accuracies were up to 60%. The correct number of endmembers to unmix every pixel was adequately identified by the iterative linear spectral unmixing. Future research efforts should be put in making use of the high revisit time of the MERIS sensor (temporal unmixing).
Since the launch of MERIS on ENVISAT long term activities using vicarious calibration approaches are set in place to monitor potential drifts in calibration in the radiance products of MERIS. We are using a stable, well monitored reference calibration site (Railroad Valley, Nevada, USA) to derive calibration uncertainties of MERIS over time. We are using interpolation of uncertainties to derive a second set of uncertainties for a national data validation in the Netherlands. A satellite image derived land use map of the Netherlands (LGN4) is used to determine the largest homogeneous land use classes using a standard purity index (SPI). Potential adjacency effects are minimized using moving window filters on the pixels of the aggregated map. Multiple error propagation is being used to assess the impact of calibration accuracy on land use classification. A classification in 9 land use classes is finally performed on MERIS FR images of the Netherlands using image based spectral unmixing and matched filtering with endmembers derived from the LGN. We conclude that the classification performance may significantly be increased, when taking into account long-term vicarious calibration results.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.