MOEA/D-UR based infrared feature extraction for hypervelocity impact spacecraft damage detection and assessment

In this paper, a decomposition-based multi-objective evolutionary algorithm updating weights when required (MOEA/D-UR) infrared feature extraction method is proposed for damage detection. Due to the complexity of the damage caused by hypervelocity impact (HVI), it is difficult to visually characterize the coupling defects in pulse thermography in reflection mode. After obtaining a multi-objective optimization problem, an adaptive metric was utilized to determine when to adjust the weights and divide the objective space. On the actual infrared data, we introduce the sample distribution objective, the data structure objective, and the local pixel relationship objective, based on which we establish a multi-objective optimization problem. The experimental results show that the detection framework is able to detect different HVI damages and realize visual evaluation. The proposed method can effectively classify point impact and surface impact damage caused by HVI. The method successfully identified 17 optimal Pareto solutions from 250 selected TTRs, using a block of 327,680 pixels. The detection framework demonstrates robust performance in practical infrared applications, precisely discriminating HVI-induced defective regions.