Machine Vision System For Automatic Crack Detection Of Iron Ore Green Pellets

Iron ore pellets are widely used in iron-making industry using blast furnaces because of its high grade and easy reduction.At present,the iron ore concentrate is generally processed by disc pelletizers into green pellets,which are then processed by sieving,roasting,hardening and cooling to become iron ore pellets(final product).Therefore,the quality of green pellets exerts a great influence on the final product quality.Drop strength is one of the important quality metrics of green pellets,and crack detection is a key step in measuring the drop strength.However,current crack detection methods depend on manual observation by human eyes,which is a bottleneck in automatic measurement of drop strength.The development of machine vision technology provides a feasible solution to the above problem.However,the current crack detection algorithms mainly focus on the flat surface of large targets such as bridges and roads,and there is no research on the spherical surface of small targets like green pellets.To solve the abovementioned problem,this paper studies a machine vision system for automatic crack detection of iron ore green pellets.The system consists of a hardware platform and a software developed for machine vision crack inspection of green pellets.The system effectively solves the bottleneck problem of human eyes observing the cracks during the drop strength measurement,which has important theoretical value and practical significance.The main contributions of the present work are as follows:(1)A hardware system for machine vision crack inspection of green pellets was designed and implemented.According to the characteristics of small size,rough surface and moisture of green pellets,a hardware platform for machine vision crack inspection of green pellets was built after determining the type of industrial camera and light source through experiments.with consideration of green pellet features.297 pellet images were captured using this platform.The analysis of the images shows that the pellet images have the characteristics of uneven illumination,local shadows,complex and changeable background.This challenges machine vision-based crack detection methods both in the measuring accuracy and in the robustness.(2)A method based on the steerable evidence filter(SEF)was proposed for crack detection of green pellet.Firstly,median filter and bilateral filter is applied to remove the image noise.Secondly,the target area of green pellets is extracted using the active contour model to eliminate the raw material stains in the image background.Thirdly,the steerable evidence filter is used to segment the area of target pellets and morphological processing is adopted to eliminate the pellet edges and noise in the response map of pellet crack,so that the pellet cracks can be well segmented.Finally,the connectivity domain method is used to detect whether there are cracks in the pellet and the number of cracks is then calculated.Results show that the proposed crack detection method has advantageous performance in detection accuracy,precision,and F1-score.The detecting accuracy of images with cracks is 96%,and the detecting accuracy for crack numbers is 90%.(3)A software for machine vision crack inspection of green pellets was developed using C#and Python programming language,with functions of camera connection,image acquisition,automatic crack detection,detection result display.The test results show that the software has friendly operation interface,fast detection speed and stable operation.The software can be put into the drop strength measurement of green pellets,complete the image collection and automatic crack detection.These functions meet the actual engineering needs.(4)The crack detection system developed in the present work was integrated with a manipulator grasping the green pellets to measure the drop strength of green pellets.Tests show that the system can replace the human eyes to automatically detect the cracks after green pellets drop several times.The system effectively solves the bottleneck problem of human eyes observing the cracks during the drop strength measurement with high accuracy,which represents the crack detection system has practical application and popularization value.