Infrared small target detection based on isolated hyperedge

Abstract

Detecting infrared small targets robustly in complex backgrounds is crucial for Infrared Search and Track (IRST) applications. However, high-intensity structures in the background, such as sharp edges, pose a challenging task, especially when the target has a low signal-to-noise ratio. We propose an Intuitionistic Fuzzy Hypergraph-based Target Detection method (IFHTD) to address this issue. IFHTD models the uncertainty of small target detection by intuitively fuzzifying the entire image at the pixel level. We define weighted intuitionistic fuzzy entropy as a membership function for target attributes in image blocks, thereby obtaining intuitionistic fuzzy sets for each image block vertex. Subsequently, the detection of infrared small targets is transformed into detecting regionally isolated hyperedges. Using intuitionistic fuzzy divergence distance metrics, we construct an intuitionistic fuzzy hypergraph for an image window. Isolated hyperedges are extracted from the centers of the image window using a predefined threshold. These isolated hyperedges are assigned weights to create a weighted graph, doubling as the infrared target’s saliency map. Experimental results demonstrate our algorithm’s robustness and effectiveness in practical infrared small target detection scenarios.