Defect extraction method based on feature mining and weighted Bayesian classifier

Abstract

The invention discloses a defect extraction method based on feature mining and a weighted Bayesian classifier, and the method comprises the steps: selecting a step length in a thermal image sequence to block an image, removing redundant information according to the block, and extracting a representative transient thermal response; then, feature extraction is carried out on the transient thermal response by utilizing a feature extraction formula; according to the extracted features and the weights of different features, obtaining a weight vector; classifying the transient thermal response by using a weighted Bayesian classifier; then, the three-dimensional matrix is transformed to obtain a two-dimensional image containing a defect area, and finally, a Canny operator is adopted to carry outedge contour extraction on the two-dimensional image containing the defect area with the largest pixel value (temperature value) difference to obtain a final defect image, so that defect characteristics of the thermal image are extracted. According to the method, the physical information contained in the transient thermal response curve is deeply mined, so that the clustering reasonability is improved, and the defect extraction accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of defect detection, and more specifically relates to a defect extraction method based on feature mining and a weighted Bayesian classifier. Background technique [0002] Infrared thermal image detection technology obtains structural information on and below the surface of the material by controlling the thermal excitation method and measuring the temperature field change on the surface of the material, so as to achieve the purpose of detection. When acquiring structural information, infrared thermal imaging cameras are often used to record the temperature field information of the surface or subsurface of the specimen over time, and convert it into a sequence of thermal images for presentation. Due to the huge amount of data and strong noise interference of the thermal image sequence obtained by the infrared thermal imager, in order to obtain better detection results, it is necessary to extract features from...

AI Technical Summary

Problems solved by technology

In this invention patent application, the fuzzy C-means algorithm classifies the transient thermal response curves through the cluster center and the membership function. It can be seen from its objective function that the classification principle is to minimize the distance between the sample and the cluster center. However, this method does not further explore the physical meaning of each transient thermal response curve

Because the physical information contained in the transient thermal response curve is not deeply excavated, the rationality of clustering is reduced, which affects the accuracy of defect extraction

Images