Research On Crack Defect Quantization Algorithm Based On ECPT Thermal Image

Non-destructive testing is to detect the structure,physical properties and state of the interior and surface of the object without damaging the object.This technology is of great significance for ensuring product quality and equipment safe service.Eddy Current Pulsed Thermography(ECPT)is an emerging transient eddy current detection technology.Because of its advantages of wide frequency spectrum,high sensitivity,fast detection speed,large detection area,and easy quantitative calculation,this technology has good applications in the nondestructive detection and evaluation of metal materials.At present,there are researches on pulsed eddy current thermal imaging detection technology,most of which are aimed at the identification and enhancement of defects in thermal images.Therefore,this article is based on the pulse eddy current thermal imaging detection technology,aimed at the surface cracks of the test piece,and explored the relationship between temperature distribution and crack depth.In addition,combined with computer vision,the crack defect boundary detection and positioning algorithm is designed from the perspective of data statistics,and then the width and length of the crack defect are detected,which provides a reference for the application of defect quantification in the industry.The research contents of this article are as follows:1)Modeling and analysis of crack defects at different depths.By modeling the inspected object and using the finite element analysis method to solve the model temperature convergence value,the influencing factors that affect the temperature field distribution on the surface of the test piece can be deeply explored.Through the comparative analysis within the group,the qualitative relationship between the depth of crack defects and the burial depth of subsurface defects and the distribution of thermal characteristics was studied.2)Design of boundary detection and location algorithm for crack defects.The crack defects on the thermal image are extracted by the thermal feature extraction method.Combined with computer vision,an algorithm that can detect and locate the boundary of crack defects in a small range is designed.The algorithm converts two-dimensional image information into one-dimensional array information through dimensionality reduction,which greatly improves the concentration of image information.Then improve the difference algorithm,by comparing the adjacent difference values,so that it can automatically find the optimal gradient value,and thus improve the sensitivity of the algorithm to the gradient value,especially the weak gradient signal.The algorithm is used to locate the range of the crack defect boundary on the pixel level,and to count the pixel information of the crack defect on the width for quantitative evaluation.3)Separation and linear abstraction of crack defects.The method of image fusion is used to eliminate redundant temperature information,retain key defect information,and improve the ability of defect detection.Helps improve the accuracy of image segmentation.In addition,through image refinement of linear features for crack defects,skeleton extraction is performed to abstract crack defects into single-pixel linear models.In this paper,through simulation and experimental research,the effect of crack defects on the surface temperature field of the inspected object is explored.At the same time,the influence of the buried depth of subsurface defects on the peak time of the surface temperature of the test piece is explained.And improve the difference algorithm,propose a boundary detection and positioning algorithm based on data statistics.Finally,a quantitative evaluation scheme for crack defect length is proposed.This scheme can accurately separate cracks from thermal images.And through the thinning algorithm of image texture to extract the skeleton of the crack defect,and then quantify the length of the defect.Finally,the algorithm in this paper is applied to thermal image processing to verify the effectiveness of the algorithm.