Semisupervised Classification With Adaptive Anchor Graph for PolSAR Images
- 8 February 2021
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Geoscience and Remote Sensing Letters
- Vol. 19 (1545598X) , 1-5
Abstract
With the rapid development of the remote sensing area, now, we can collect massive data. Labeling polarimetric synthetic aperture radar (PolSAR) data is a labor-intensive and time-consuming work. The performance of the supervised model heavily depends on the number of training samples and there is a large number of unlabeled data unutilized, which contains statistical information of the PolSAR image. To solve this issue, in this letter, an adaptive anchor graph regularization (AdaAGR) is proposed for PolSAR images. The accuracy of the conventional anchor graph-based method is dependent on the choice of the number of nearest anchors. Applying a global number of nearest anchors to the anchor graph method ignores the local data density. Low data density corresponds to a small number of nearest anchors and vice versa. A global number of nearest anchors may limit the performance of classification. In this letter, an adaptive anchor graph strategy is proposed. First, cluster centers generating from Wishart clustering are served as graph anchors. Second, the distances between the data point and its closest anchors are calculated in ascending order. Then, an optimal number of nearest anchors is selected for every pixel. Third, the similarity between data and its neighboring anchors is measured, and then, an anchor graph is constructed. Finally, the anchor graph is regularized and the unlabeled samples are predicted. Experimental results on two real PolSAR data sets show that the classification accuracy can be further improved by employing the proposed adaptive anchor graph strategy.Keywords
Funding Information
- National Natural Science Foundation of China (61801098)
- National Natural Science Foundation of China (61971101)
- Science and Technology department of Yibin (2019GY001)
This publication has 15 references indexed in Scilit:
- Terrain classification based on spatial multi-attribute graph using Polarimetric SAR dataApplied Soft Computing, 2018
- Feature guided Gaussian mixture model with semi-supervised EM and local geometric constraint for retinal image registrationInformation Sciences, 2017
- Scalable Semi-Supervised Learning by Efficient Anchor Graph RegularizationIEEE Transactions on Knowledge and Data Engineering, 2016
- Large Polarimetric SAR Data Semi-Supervised Classification With Spatial-Anchor GraphIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2016
- Semi-supervised classification via kernel low-rank representation graphKnowledge-Based Systems, 2014
- Semi-supervised classification based on subspace sparse representationKnowledge and Information Systems, 2013
- Support Vector Machine for Multifrequency SAR Polarimetric Data ClassificationIEEE Transactions on Geoscience and Remote Sensing, 2009
- Unsupervised classification of polarimetric synthetic aperture Radar images using fuzzy clustering and EM clusteringIEEE Transactions on Geoscience and Remote Sensing, 2005
- Unsupervised classification using polarimetric decomposition and the complex Wishart classifierIEEE Transactions on Geoscience and Remote Sensing, 1999
- An entropy based classification scheme for land applications of polarimetric SARIEEE Transactions on Geoscience and Remote Sensing, 1997