Identification And Rating Of Shrinkage Cavity And Crack In The Continuous Casting Square Slab Based On Image Processing

Detection of continuous casting billet's macrostructure is an important link in the whole detection process of steel quality.At present,the detection of the continuous casting billet low-times defects is still relying on manual rating.This method has the shortcomings of subjective criteria inconsistent and low detection efficiency.In order to improve the efficiency of defect detection and rating accuracy,the thesis used image processing technology as a way of defect identification and rating.The national standard provides a variety of continuous slab low-time defects,but cracks and shrinkage cavity are the most common and the most harmful defects,so the thesis select crack and shrinkage as the object of study.The main research contents are as follows:(1)Classification of the rating parameters of the continuous casting slab low-times defects:crack and shrinkage cavity.First,the thesis analysed of the national standard:YB/T 4002-2013?Rating Chart of the Continuous Casting Slab Low-times Defects?,and then identified the characteristic parameters of crack and shrinkage cavity.(2)Segmentation,feature selection,recognition and rating of shrinkage cavities.According to the location and the grayscale characteristics of the shrinkage cavity,the center of the slab was cut off for reduce the research area.Then used the threshold segmentation method based on the gray mean value to separate shrinkage cavity from the background.Used the morphological transformation,connectivity region detection and other image processing technology on the binary image to remove small interference and large cracks,and remained the suspected shrinkage cavity area only.Extracted circularity,difference between the foreground and background average value,standard deviation,average gradient and the gradient of the edge of the suspected shrinkage cavity area to identify the shrinkage cavity.Finally,the shrinkage cavities were rated referring to the national standard.(3)Segmentation,feature selection,recognition and rating of cracks.According to the location,grayscale and shape characteristics,the thesis adopted the twice detection method to improve the detection accuracy.In the first detection,used the threshold segmentation method based on the most gray value to separate cracks from the background roughly.And then repaired the fractured cracks and remove the interference.Extracted the circularity,average gradient,energy,entropy,and inverse gap of the suspected crack area to identify cracks.Detected the identified crack area again and abstracted the number and the length of cracks.Finally,cracks were rated referring to the national standard.(4)Preparation of the software.Wrote the human-computer interaction interface,and combined it with the image processing program to finish a software.The software function had the detection of continuous casting slab's macrostructure defects,exhibition of the national standard,management of results and production parameters,and system settings.168 continuous casting slab macrostructure pictures of shrinkage cavity were processed,shrinkage cavity's identification successful rate was up to 90%,and the rating successful rate was up to 84.5%.112 continuous casting slab macrostructure pictures of crack were processed,crack's identification successful rate was up to 90%,and the rating successful rate was up to 83.0%.