Study On Defect Inspection Method In Hot Continuous Casting Slab Surface Based On Allogenetic CCD Laser Scan Imaging And Information Fusion

Continuous casting (CC) technology in iron and steel enterprises is the miansolification process for molten steel and has been widely applied. The surface quality ofthe CC production determines the process operation of the production line, meanwhile,it is an important factor that influences the metal yield and the cost, what is more, it isalso a premise condition for the CC product hot charging (CC-HC) and direct rolling(CC-DR) technology. In general, it is very challenging to guarantee the defect-free CCproduct just by improving the production process technology, therefore, it is a feasibleway by removing the defect CC product by surface quality assessment system toimprove the final strip quality and reduce the cost that means too much to the modernindustry.The Charge-coupled Device(CCD) based image scanning technology(line scanningand area scanning) for defect inspection is a burgeoning nondestructive detectionstrategy, which has become a research hotspot in CC product defect inspection filed as aresult of its particular advantages such that non-contact feature and fast responseperformance and high signal-to-noise ratio and strong anti-jiamming capability. Basedon the high-definition scanned images the imaging processing and pattern recognitionand ROI retrieval and quantitative analysis strategies can be applied into the defectanalysis.Therefore, In order to realize a reliable detection system for CC slab surfacedefects inspection, based on the CCD scan imaging techniques, this paper carried outthe following research contents and the corresponding findings are also presented:①In allusion to automatically remove the fake defects information like thesurface scales that its imaging features are very similar with the real defects, the CCDimaging techniques must be researched into at high noise background to acquire reliableimages. Therefore, in the imaging strategies in system based on the active mode, amaximum information entropy algorithm was established for acquiring the highdefinition image in linear scan CCD images and a Tenengrad based laser stripe optimalimaging method was proposed for extracting the laser stripe precisely in3D area scanCCD images.②To provide an accurate surface defects inspection system and make theautomation of robust image delineation method a reality in routine CC production line, a general approach is presented for continuous casting slab (CC-slab) surface defectsextraction and delineation. The applicability of the system is not tied to CC-slabexclusively. We combined the line array CCD traditional scan imaging (LSI) and areaarray CCD laser three-dimensional (3D) scan imaging (ASI) strategies in designing thesystem. Its aim is to suppress the respective imaging system’s limitations. In the system,the images acquired from the two CCD sensors are carefully aligned in space and intime by maximum mutual information (MMI)-based full-fledged registration schema.Subsequently, the image information from these two subsystems such as the unbroken2D information in LSI and3D depressed information in ASI is fused.③To improve the2D ROI region segmentation accuracy, the image informationfrom these two subsystems such as the unbroken2D information in LSI and3Ddepressed information in ASI is fused. On the basis of the established dual CCD scanimaging system and rough set theory the ROI localization by seed specification wasdevised, and meanwhile, the foreground seed sets and background seed sets and thefuzzy reclassification region were defined, finally, the delineation for ROI by theiterative relative fuzzy connectedness (IRFC) algorithm was utilized to get the preciseinspection result.④Through analyzing the characteristics of CCD imaging model in ASIsubsystem and based on experimentally scanning measurement study, we establishedthe image depth information quantitative detection method by laser stripe modulation,which a devised2D calibration template was utilized, its aim is to provide reliable datainformation for feature fusion in system. Meanwhile, on the basis of fusion results of2D feature in LSI and3D information in ASI, the techniques of surface3D shapereconstruction method was studied. The calibration for the system is based on theproposed system optimal imaging strategy. The proposed calibration algorithm tested inthis paper has a promise precision and a good ability of resistance to external lightinterference.⑤Through analyzing the existing problems in CC product surface for machinevision based defect detection method, the algorithm proposed in this paper wasimplemented by software development and improved based on the field experiments. Inorder to optimally determine the inspection schema the oscillation mark interference inimages was taken into account. On the other hand, aiming to the fake noisy points inimages a morphological processing algorithm was proposed for correcting the seedspecification. Finally, the horizontal and vertical interval maximum value determination method was devised for judging the defect depth and oscillation mark that utilize thelocal analyzing method to recognize the defects within the wave valley.The paper was sponsored by National Natural Science Foundation of China andShanghai Bao Steel Group Co. United Research Foundation (50974151) named by“Multi-senor based hot slab surface defect inspection in continuous casting process”.Based on the allogenetic CCD laser scanning imaging strategy, the surface defectinspection method was profoundly researched for CC slab in high noise background,meanwhile, a series of experiments and on-site verification were implemented. Theproposed laser scanning imaging detection techniques have a good promise.