Dataset Information

Integrating local and global error statistics for multi-scale RBF network training: an assessment on remote sensing data.

ABSTRACT:

Background

This study discusses the theoretical underpinnings of a novel multi-scale radial basis function (MSRBF) neural network along with its application to classification and regression tasks in remote sensing. The novelty of the proposed MSRBF network relies on the integration of both local and global error statistics in the node selection process.

Methodology and principal findings

The method was tested on a binary classification task, detection of impervious surfaces using a Landsat satellite image, and a regression problem, simulation of waveform LiDAR data. In the classification scenario, results indicate that the MSRBF is superior to existing radial basis function and back propagation neural networks in terms of obtained classification accuracy and training-testing consistency, especially for smaller datasets. The latter is especially important as reference data acquisition is always an issue in remote sensing applications. In the regression case, MSRBF provided improved accuracy and consistency when contrasted with a multi kernel RBF network.

Conclusion and significance

Results highlight the potential of a novel training methodology that is not restricted to a specific algorithmic type, therefore significantly advancing machine learning algorithms for classification and regression tasks. The MSRBF is expected to find numerous applications within and outside the remote sensing field.

SUBMITTER: Mountrakis G

PROVIDER: S-EPMC3411665 | biostudies-literature | 2012

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Integrating local and global error statistics for multi-scale RBF network training: an assessment on remote sensing data.

Mountrakis Giorgos G Zhuang Wei W

PloS one 20120802 8

<h4>Background</h4>This study discusses the theoretical underpinnings of a novel multi-scale radial basis function (MSRBF) neural network along with its application to classification and regression tasks in remote sensing. The novelty of the proposed MSRBF network relies on the integration of both local and global error statistics in the node selection process.<h4>Methodology and principal findings</h4>The method was tested on a binary classification task, detection of impervious surfaces using ...[more]

PMID: 22876278

Similar Datasets

Project description:Monitoring ecosystem functioning is a significant step towards detecting changes in ecosystem attributes that could be linked to land degradation and desertification in drylands worldwide. Remote sensing-based vegetation indices (VIs) and land surface albedo are two favorite indicators to monitor desertification process due to their close relationship with ecosystem status and to their increasing applicability over multiple spatiotemporal scales. While VIs are routinely used to monitor ecosystem attributes and functions such as vegetation cover and productivity, no previous study has evaluated whether remote sensing-measured albedo is related to the simultaneous provision of multiple ecosystem functions (multifunctionality) in global drylands. In this study, we evaluated the correlation of six albedo metrics (shortwave black-sky albedo, shortwave white-sky albedo, visible black-sky albedo, visible white-sky albedo, near-infrared black-sky albedo and near-infrared white-sky albedo) and two VIs (Normalized Difference Vegetation Index (NDVI) and Enhanced Vegetation Index (EVI)) with multifunctionality indices related to carbon, nitrogen and phosphorus cycling measured in 61 dryland ecosystems from all continents except Antarctica. We found a negative relationship between land surface albedo and multifunctionality. Black-sky albedo had a stronger correlation with multifunctionality than white-sky albedo. Visible black-sky albedo showed the strongest correlation with multifunctionality (MUL, -0.314), as well as with functions related to carbon (CCY, -0.216) and nitrogen cycling (NCY, -0.410), while near-infrared (-0.339) and shortwave black-sky albedo (-0.325) showed stronger correlations with functions related to phosphorus cycling (PCY) than visible black-sky albedo (-0.233) did. VIs showed significant positive correlations with MUL, CCY, and NCY, and the magnitudes were higher than those observed between albedo metrics and the multifunctionality indices. However, VIs were not correlated with PCY, which had significant correlations with both shortwave and near-infrared albedo. Though the magnitudes of the correlations observed were not high, which may result from the wide variability in soil and vegetation types in our dataset, our findings indicate that remotely sensed albedo correlates to multifunctionality, which has been linked to alternative states in global drylands. As such, albedo has the potential to monitor changes in dryland ecosystem functioning, which can inform us about the onset of desertification in these areas.

Dataset Information

Integrating local and global error statistics for multi-scale RBF network training: an assessment on remote sensing data.

Background

Methodology and principal findings

Conclusion and significance

Publications

Integrating local and global error statistics for multi-scale RBF network training: an assessment on remote sensing data.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure