Abstract
For large-scale mapping applications, cross-domain hyperspectral image classification (HSIC) has emerged as a highly promising research area. However, the classification accuracy decreased significantly when unseen classes emerged. Few shot learning (FSL) methods are adopted in cross-domain HSIC methods to address this problem. Despite this, existing cross-domain HSIC methods still have three key issues that hamper their classification capabilities: 1) Previous works struggle to balance incorporating distinctive intra-domain knowledge and managing model complexity in the face of significant domain representation differences, 2) Previous works inadequately consider the limited capture capacity of inter-domain intrinsic mutually-invariant structures, and 3) Previous works fail to capture the distinct characteristics of both head categories (e.g., urban buildings) and tail categories (e.g., urban corn) simultaneously when apply FSL to deal with unseen classes problem. In this paper, we propose a cycle-resemblance few-shot transformation (CF-Trans) network to effectively handles the aforementioned challenges by integrating intra-domain distinctiveness with inter-domain invariance. To facilitate efficient intra-domain feature aggregation for HSI, a novel lightweight intra-domain attentive network is introduced. Different from previous works, to reduce the negative impact caused by inaccurate classifier predictions, from the perspective of inter-domain knowledge transformation, a cycle-resemblance adversarial network is designed to capture the intrinsic mutually-invariant structures. A dynamic label expansion mechanism is designed to capture the distinctive intra-domain features of the head and tail classes. Experimental results on six HSI datasets including agricultural, rural-urban and urban datasets show the remarkably performance of our network.
Funding Information
  • National Key Research and Development Program of China (2022YFB3903402)
  • National Natural Science Foundation of China (42271413)
  • Open Fund of State Key Laboratory of Remote Sensing Science (OFSLRSS202329)

This publication has 63 references indexed in Scilit: