Pseudo-Labelling Contrastive Learning for Semi-supervised Hyperspectral and LiDAR Data Classification

Elevation information from light detection and ranging data (LiDAR) relieve the phenomenon of “same spectrum with different object” in Hyperspectral images (HSI) classification. Consequently, HSI and LiDAR joint classification is a significant research topic. However, existing methods encounter several challenges. Primarily, there exists a deficiency in intra-class information interaction and underutilization of discriminative feature. Furthermore, the process of labeling samples is time-consuming and laborious. To solve the aforementioned issues, a classification method based on pseudo-labelled contrastive learning is proposed to exploit substantial amounts of unlabelled information in order to enhance intra-class information interaction. The proposed method is divided into two stages for semi-supervised classification. In the first stage, an unsupervised feature extraction network is designed to improve the interaction of features from multi-modal data. A multi-modal data cross-attention (MCA) module is proposed to enhance the interaction of multi-modal information at corresponding locations. Exploiting pseudo-labelling contrastive learning (PLC) module facilitates the interaction of information between intra-class objects. In the second stage, supervised classification with a limited number of labeled samples is performed. The multi-source discriminatively consolidate feature (MDCF) module is designed to generate discriminative features, which are used to guide the fusion feature enhancement process. Apart from this, this module leverages multi-scale features to expand the receptive field. Tested on both self-constructed and public datasets, the proposed method provides higher classification accuracy than some existing methods with a limited amount of labeled samples.