Research On Automatic Defect Detection Identification And Evaluation Method Based On Infrared Thermal Wave

Infrared non-destructive testing techniques are used in a wide range of fields to detect defects quickly and visually without damaging the material itself.Automatic detection and identification of defects is critical to the operational safety of systems,especially in aerospace,where complex multi-type defects formed by hypervelocity impacts pose further difficulties for detection.Defect features between different types may have similarity leading to difficulty in differentiation and may be lost due to complex detection background.Based on this,this thesis designs an automatic defect detection and identification algorithm based on infrared detection technology and proposes an evaluation method to assess the overall defect detection capability of the detection scheme.The main research elements of this thesis are:1.For the problem that complex defect types are difficult to distinguish,a method for automatic discrimination of defect features is proposed.Considering the possible simultaneous occurrence of many defect categories and complex distribution,this thesis proposes an adaptive density peak clustering algorithm,which adaptively determines the clustering center according to the distribution of the evaluation function,after which the IR features are labeled as cluster core,cluster backbone and boundary points respectively according to the local density,and the labels of the clustering center are passed outward in turn,and finally the defect feature classification results are obtained to achieve automatic determination of the number of defect categories.2.In order to meet different inspection requirements,a quantitative defect detection method is designed,which combines spatial location information and physical feature quantification information to describe defect characteristics.The accurate defect location results can not only help to quickly find the defect form and location,but also guide the secondary local detection of tiny defects,thus improving the detection efficiency of defects.Based on this,the method takes into account that the initially extracted defect features may have the problem of misclassification of pixel points in the defect area,and the defect feature images are processed before the defect analysis.After that,the coarse localization of the whole pixel edges of the defect is performed by the boundary tracking method,and then the sub-pixel contour localization of the defect is achieved by applying the boundary contour fitting algorithm.Finally,defect quantification was performed to visualize the defect feature information and provide specific numerical reference for the damage of the specimen.3.In order to measure the detection performance of the inspection scheme for complex specimens and to analyze and describe and evaluate the defect detection in different inspection sessions,this thesis designs a method for evaluating the defect detection capability of an infrared inspection scheme.Considering that the quantitative calculation results of defects are based on purely image information statistics on the basis of the defect feature recognition images obtained by image processing.Therefore,this thesis uses the extracted image defect characterization capability as a carrier to measure the detection capability of both infrared sequence processing technology and image processing technology in the infrared detection scheme by combining several defect image evaluation indexes at the same time.Finally,the evaluation results of the two aspects are combined to evaluate the overall defect detection capability of the IR detection scheme.