Semi-Supervised Classification of PolSAR Images Based on Co-Training of CNN and SVM with Limited Labeled Samples

Abstract

Recently, convolutional neural networks (CNNs) have shown significant advantages in the tasks of image classification; however, these usually require a large number of labeled samples for training. In practice, it is difficult and costly to obtain sufficient labeled samples of polarimetric synthetic aperture radar (PolSAR) images. To address this problem, we propose a novel semi-supervised classification method for PolSAR images in this paper, using the co-training of CNN and a support vector machine (SVM). In our co-training method, an eight-layer CNN with residual network (ResNet) architecture is designed as the primary classifier, and an SVM is used as the auxiliary classifier. In particular, the SVM is used to enhance the performance of our algorithm in the case of limited labeled samples. In our method, more and more pseudo-labeled samples are iteratively yielded for training through a two-stage co-training of CNN and SVM, which gradually improves the performance of the two classifiers. The trained CNN is employed as the final classifier due to its strong classification capability with enough samples. We carried out experiments on two C-band airborne PolSAR images acquired by the AIRSAR systems and an L-band spaceborne PolSAR image acquired by the GaoFen-3 system. The experimental results demonstrate that the proposed method can effectively integrate the complementary advantages of SVM and CNN, providing overall classification accuracy of more than 97%, 96% and 93% with limited labeled samples (10 samples per class) for the above three images, respectively, which is superior to the state-of-the-art semi-supervised methods for PolSAR image classification.

Land cover analysis of PolSAR images using probabilistic voting ensemble and integrated support vector machine

“…We decided to compare the results with state-of-the-art PolSAR image classification utilizing supervised methodologies used in the most recent research studies. We select two fully supervised methods, FS-SVM and FS-CNN, and one self-training method ST-SVM, used in the reference paper 13 to compare with our proposed method. Also, the classical classification results of each target decomposition type are utilized in the comparison.…”

“…These CNN-based models have demonstrated remarkable capabilities in various classification tasks and have been extended and enhanced for PolSAR image classification 9 , 10 . Several papers and research studies have addressed similar issues and have contributed to the field, for example, semi-supervised complex-valued generative adversarial networks, 11 classification of polarimetric SAR images using compact CNNs, 12 and semi-supervised classification of PolSAR images based on co-training of CNN and SVM 13 . These advancements in research are driving progress in PolSAR image classification and opening up new avenues for more accurate and practical remote sensing applications.…”

“…9,10 Several papers and research studies have addressed similar issues and have contributed to the field, for example, semi-supervised complex-valued generative adversarial networks, 11 classification of polarimetric SAR images using compact CNNs, 12 and semisupervised classification of PolSAR images based on co-training of CNN and SVM. 13 These advancements in research are driving progress in PolSAR image classification and opening up new avenues for more accurate and practical remote sensing applications. However, it is worth noting that CNNs are data-driven networks that typically require many labeled samples for effective training.…”

“…Fig 13. Classification maps of the San-Francisco dataset given by six decomposition methods in the first column, with the ground truth information overlaid in the second column, and the class label prediction probability layer in the third column depicted using the jet color map and ranging from [0:1].…”

PolSAR Image Classification Via a Multigranularity Hybrid CNN-ViT Model With External Tokens and Cross-Attention

“…Since CNNs have strong local feature extraction capability and have been employed as feature extractors for PolSAR images in many works [43]- [46], a multilayer CNN is designed for extracting local features from PolSAR images. The CNN involves three blocks consisting of convolutional and pooling layers.…”