A decomposition multi-objective optimization method based on dynamic weight adjustment for infrared thermal image defect feature adaptive extraction method

摘要

In order to visually display defect features, it is necessary to extract features from infrared thermal sequence data. Due to the different types of pixels in the damaged region are motivated by the heat after the temperature change is different, according to the transient thermal responses (TTRs) of pixels change trend, with the improved mean-shift clustering algorithm relies on pixel itself in the space the temperature characteristic of adaptive partition for defect type, obtain the TTRs of different types of damage. In typical TTRs of extracted from each damage set, considering the similarity of similar pixel point temperature and similar pixel point temperature difference, choose the multi-objective optimization algorithms, but may occur in the process of solving uneven Pareto front led to the decrease of the diversity of solution set, so here use a dynamic weighting multi-objective optimization adjustment of decomposition algorithm from the various damage focus on selected typical TTRs.