Scope

The Image Analysis and Classification section of Frontiers in Remote Sensing seeks to publish original research covering all aspects of remote sensing image analysis. Spanning the full spectrum from physical characterization and model inversion to thematic classification and machine learning application. The emphasis of this section is on rigorous, repeatable, physical and quantitative approaches to image analysis. Specific topics include, but are not limited to:

· Physical property retrieval (e.g. surface temperature & emissivity, surface reflectance, soil moisture estimation, evapotranspiration, ocean color and fluorescence)

· Land cover characterization and change detection (e.g. spectral mixture analysis, vegetation abundance & structure, water quality)

· Hyperspectral image analysis (e.g. spectrometer characterization & calibration, BRDF correction, intrinsic dimensionality, anomaly detection)

· Spatio-Temporal analysis (e.g. exploratory data analysis, dimensionality reduction, temporal mixture modeling, Hierarchical Dynamical Spatio-Temporal models)

· Multi-Sensor analysis (e.g. spatial-spectral fusion and sharpening, cross-sensor harmonization, multi-scale interpolation, multimodal biophysical characterization)

· Mapping and monitoring approaches (e.g. vegetation phenology mapping, agriculture & rangeland monitoring, forest cover mapping, fire detection, settlement mapping)

· Thematic classification (feature space characterization, spectral separability, model construction, interpretable machine learning, model-agnostic interpretation)

Publishable work must present analyses in the context of the current state of knowledge with full attribution of prior related work. Retrieval and abundance estimation methodologies must be physically-based and field or vicariously validated. All proposed methodologies must demonstrate sufficient generality of application to preclude geographic specificity. Classification and machine learning approaches must specify theoretical or statistical bases or sensitivity analyses to justify hyperparameter selection and must address potential overfitting explicitly. All experimental analyses must provide sufficient detail as to be repeatable or replicable.

Submission

Copyright Statement