MAPM:PolSAR Image Classification with Masked Autoencoder Based on Position Prediction and Memory Tokens
Open Access
- 17 November 2024
- journal article
- research article
- Published by MDPI AG in Remote Sensing
- Vol. 16 (22) , 4280
Abstract
Deep learning methods have shown significant advantages in polarimetric synthetic aperture radar (PolSAR) image classification. However, their performances rely on a large number of labeled data. To alleviate this problem, this paper proposes a PolSAR image classification method with a Masked Autoencoder based on Position prediction and Memory tokens (MAPM). First, MAPM designs a Masked Autoencoder (MAE) based on the transformer for pre-training, which can boost feature learning and improve classification results based on the number of labeled samples. Secondly, since the transformer is relatively insensitive to the order of the input tokens, a position prediction strategy is introduced in the encoder part of the MAE. It can effectively capture subtle differences and discriminate complex, blurry boundaries in PolSAR images. In the fine-tuning stage, the addition of learnable memory tokens can improve classification performance. In addition, loss is used for MAE optimization to enhance the robustness of the model to outliers in PolSAR data. Experimental results show the effectiveness and advantages of the proposed MAPM in PolSAR image classification. Specifically, MAPM achieves performance gains of about 1% in classification accuracy compared with existing methods.
Keywords
Funding Information
- National Natural Science Foundation of China (62201201, 62201407, B2022-15, 2022M722496)
- Doctoral Foundation of Henan Polytechnic University (62201201, 62201407, B2022-15, 2022M722496)
- China Postdoctoral Science Foundation (62201201, 62201407, B2022-15, 2022M722496)
This publication has 52 references indexed in Scilit:
- Land use and land cover classification with hyperspectral data: A comprehensive review of methods, challenges and future directionsNeurocomputing, 2023
- Gaussian-type activation function with learnable parameters in complex-valued convolutional neural network and its application for PolSAR classificationNeurocomputing, 2022
- The encoding method of position embeddings in vision transformerJournal of Visual Communication and Image Representation, 2022
- PolSAR image classification via a novel semi-supervised recurrent complex-valued convolution neural networkNeurocomputing, 2020
- PolSAR image classification based on multi-scale stacked sparse autoencoderNeurocomputing, 2019
- Building damage assessment from PolSAR data using texture parameters of statistical modelComputers & Geosciences, 2018
- Fully PolSAR image classification using machine learning techniques and reaction-diffusion systemsNeurocomputing, 2017
- Adaptive land classification and new class generation by unsupervised double-stage learning in Poincare sphere space for polarimetric synthetic aperture radarsNeurocomputing, 2017
- Stacked Convolutional Auto-Encoders for Hierarchical Feature Extraction2011
- New decomposition of the radar target scattering matrixElectronics Letters, 1990