Abstract:Spectral-spatial processing has been increasingly explored in remote sensing hyperspectral image classification. While extensive studies have focused on developing methods to improve the classification accuracy, experimental setting and design for method evaluation have drawn little attention. In the scope of supervised classification, we find that traditional experimental designs for spectral processing are often improperly used in the spectral-spatial processing context, leading to unfair or biased performance evaluation. This is especially the case when training and testing samples are randomly drawn from the same image—a practice that has been commonly adopted in the experiments. Under such setting, the dependence caused by overlap between the training and testing samples may be artificially enhanced by some spatial information processing methods, such as spatial filtering and morphological operation. Such enhancement of dependence in return amplifies the classification accuracy, leading to an improper evaluation of spectral-spatial classification techniques. Therefore, the widely adopted pixel-based random sampling strategy is not always suitable to evaluate spectral-spatial classification algorithms, because it is difficult to determine whether the improvement of classification accuracy is caused by incorporating spatial information into classifier or by increasing the overlap between training and testing samples. To tackle this problem, we propose a novel controlled random sampling strategy for spectral-spatial methods. It can greatly reduce the overlap between training and testing samples and provides more objective and accurate evaluation.

Weighted Combination Random Classifier of High Spectral Remote Sensing Image:Design and Application

An Integrated Classification Method for Thematic Mapper Imagery of Plain and Highland Terrains

Classification of Remote Sensing Image of Beijing-1 Micro-satellite Based on Multiple Classifier Combination

A Comparison of Multiple Classifier Combinations Using Different Voting-Weights for Remote Sensing Image Classification

On the Sampling Strategy for Evaluation of Spectral-Spatial Methods in Hyperspectral Image Classification

On the Sampling Strategies for Evaluation of Joint Spectral-Spatial Information Based Classifiers

Multiple Classifier Combination for Hyperspectral Remote Sensing Image Classification

A New Robust Approach for Remote Sensing Image Regional Classification

Study on the Design of Remote Image Classifier with HighResolution and Multi-spectrum Based on Neural Net

Multispectral Image Classification Based on Improved Weighted MRF Bayesian

Target Identification from High Resolution Remote Sensing Image Based on Multiple Classifier Combination

A New Kind of Combination Classifier and Its Application in Classification of Remote Sensing Image

Combined Multi-Kernel Support Vector Machine and Wavelet Analysis for Hyperspectral Remote Sensing Image Classification

Key Techniques and Implementation of Multiple Classifier System for Remote Sensing Images

Target Identification from High Resolution Remote Sensing Image by Combining Multiple Classifiers

Classification of Hyperspectral Imagery Based on Spectral Gradient, SVM and Spatial Random Forest

The Study of Object-Oriented Classification Method of Remote Sensing Image

The Remote Sensing Image Spectrum Simulation with Mixed Pixel

A Complementary Spectral-Spatial Method for Hyperspectral Image Classification

The geographical weighted K-NN classifiers in land cover classification from remote sensing image: A case study of a subregion of Xi'an, China

Classification of High-Spatial Resolution Imagery Based on Distance-Weighted Markov Random Field with an Improved Iterated Conditional Mode Method