Defect Detection And Application Of High Temperature Furnace Tile Based On Infrared Thermography

In chemical production,a large number of high-temperature furnaces will be used,and in the process of working high-temperature furnaces,crack damage will occur inside the furnace tile due to heating and high pressure,and the internal defects of the object cannot be detected and identified by the naked eye.In order to study the method of high temperature shingle defect detection,this paper collects data sets by infrared thermography,and uses morphology and neural network to study the high-temperature shingle defect detection,and the main research contents are as follows:1.In this paper,aiming at the research on the defect detection of high-temperature furnace tiles,the original sample pictures of high-temperature furnace tiles were taken by infrared thermography,and the data set of various defects such as star points,cracks,cavities and other defects of high-temperature furnace tiles was made as the main data set of this study,and the dataset was expanded by image enhancement,image stitching and other methods,and the expanded 8000 pictures were adaptively filtered and labeled.2.In terms of morphological methods,this paper compares some classical image segmentation algorithms,and finally selects the maximum inter-class variance method to select the optimal threshold.In the binarized images generated by the optimal threshold,some of the images have problems such as missing feature points,feature fractures and isolated noise.In this paper,adaptive morphological filtering is designed to process the binarized images,which can effectively alleviate the above problems.After filtering,Canny operator is used to extract image edges and calculate morphological features such as defect edge length and defect contour area,analyze the morphological feature data and design the classification scheme,which has a good classification effect.3.In terms of deep learning method,by unifying parameter settings,using accuracy and recall evaluation standards,the defect detection performance of Faster-rcnn,YOLOV3-Tiny and YOLOV4-Tiny network models was compared,and the Yolov4-tin network model with high accuracy was selected to improve,and the two identical Residual Networks were combined with the idea of image feature fusion in view of the characteristics of large scale change of high-temperature furnace tile dataset Blocks are combined and added to the Res Blocks-D module,Channel Attention and spatial Attention is used to extract the characteristic information of effective detection of objects extracted additional global and local characteristics,improve detection accuracy and speed,tested the enhanced YOLOV4-tiny compared to the original model YOLOV4-tiny model map value of 1.8%.4.Design and implement the visualization program of high temperature furnace tile defect detection,and integrate some algorithms in this paper into the program to improve the efficiency of detection.In summary,this paper analyzes the infrared defect detection of high-temperature furnace tiles by combining morphological methods and deep learning methods,and obtains good detection and classification effects.