Infrared image reconstruction for hypervelocity impact damage inspection in spacecraft using weight-adaptive multi-objective feature selection and adaptive classification

To address the challenge of detecting complex damage caused by hypervelocity impacts during spacecraft on-orbit maintenance, this paper proposes an infrared defect image reconstruction method via dynamic multi-objective feature selection. First, this study employs an adaptive classification algorithm driven by transient thermal response (TTR) density information to achieve unsupervised and precise separation of damage types. Next, a dual-objective optimisation model is developed to balance intra-class representativeness and inter-class differences. By integrating an improved decomposition-based multi-objective evolutionary algorithm (MOEA/D) with a dynamic weight vector adjustment strategy, the method optimises the selection of typical TTRs under an irregular Pareto front. Finally, the extracted high-quality TTRs are used to generate the reconstructed infrared images. Experimental results demonstrate that this method effectively highlights distinct defect features, enhances image discriminability, and significantly improves the ability to detect and extract complex damage.