Classification of hyperspectral images using self-training and a pseudo validation set

ABSTRACT It is difficult to acquire sufficient labeled samples to classify remote sensing images. Supervised classification of hyperspectral remote sensing images does not perform well in the absence of an adequate amount of labeled samples of good quality. Self-training is a popular semi-supervised approach to train a classifier using fewer labeled samples and a pool of large unlabeled samples in an iterative manner. However, the traditional self-training approach is designed to use only confident samples. This criteria need not improve the accuracy of the classifier. In this work, the use of informative and confident samples for self-training has been proposed. The technique to selects informative and confident samples using pseudo validation set is a novel attempt to improve the scope of self-training. The pseudo validation set has been extracted from a freshly classified pool of pixels based on neighborhood information. The experimental results with popular benchmark hyperspectral images show that the proposed method has the potential to perform well.

[1]  Xiaojin Zhu,et al.  --1 CONTENTS , 2006 .

[2]  HerreraFrancisco,et al.  Self-labeled techniques for semi-supervised learning , 2015 .

[3]  Dan Hu,et al.  Semisupervised Hyperspectral Image Classification Based on Generative Adversarial Networks , 2018, IEEE Geoscience and Remote Sensing Letters.

[4]  Francisco Herrera,et al.  Self-labeled techniques for semi-supervised learning: taxonomy, software and empirical study , 2015, Knowledge and Information Systems.

[5]  Paul Geladi,et al.  Techniques and applications of hyperspectral image analysis , 2007 .

[6]  Lorenzo Bruzzone,et al.  A Novel Transductive SVM for Semisupervised Classification of Remote-Sensing Images , 2006, IEEE Transactions on Geoscience and Remote Sensing.

[7]  Jon Atli Benediktsson,et al.  Semisupervised Self-Learning for Hyperspectral Image Classification , 2013, IEEE Transactions on Geoscience and Remote Sensing.

[8]  Jun Li,et al.  A novel semi-supervised hyperspectral image classification approach based on spatial neighborhood information and classifier combination , 2015 .

[9]  David A. Clausi,et al.  ST-IRGS: A Region-Based Self-Training Algorithm Applied to Hyperspectral Image Classification and Segmentation , 2018, IEEE Transactions on Geoscience and Remote Sensing.

[10]  Ye Zhang,et al.  A Novel Synergetic Classification Approach for Hyperspectral and Panchromatic Images Based on Self-Learning , 2016, IEEE Transactions on Geoscience and Remote Sensing.

[11]  Lorenzo Bruzzone,et al.  Active and Semisupervised Learning for the Classification of Remote Sensing Images , 2014, IEEE Transactions on Geoscience and Remote Sensing.

[12]  Ujjwal Maulik,et al.  A self-trained ensemble with semisupervised SVM: An application to pixel classification of remote sensing imagery , 2011, Pattern Recognit..