Research On Image Processing Methods For Strip Steel Surface Defect Based On Swarm Intelligent Optimization Algorithm

Strip steel is widely used in machinery manufacturing,automobile industry,aerospace,instrumentation,etc.In recent years,with the upgrading and transformation of production equipment,the dimensional accuracy,shape accuracy and mechanical property of strip steel have been well controlled.However,quality accidents caused by strip steel surface defect have occurred from time to time,and this causes enterprises suffer serious economic losses.The machine vision technology can realize the rapid and comprehensive inspection for strip steel surface defect,and has become an important means to ensure the surface quality of strip steel.As the core link of this technology,the performance of the image processing methods for strip steel surface defect directly affects the final detection results,so it has attracted great attention from the related scholars and engineering technicians.At present,the swarm intelligence optimization algorithm has been successfully applied to the image processing for strip steel surface defect,but there are some problems,such as the exploration capability is poor,the convergence speed is slow,and the searching accuracy is low,which affects the image processing effect.Therefore,this dissertation in-depth and systematically analyses to ant lion optimizer(ALO),bat algorithm(BA)and grasshopper optimization algorithm(GOA),and proposes improved ant lion optimizer(IALO),adaptive bat algorithm(ABA)and grasshopper optimization algorithm with principal component analysis(PCA-GOA).Then,apply the algorithms to the fields of image enhancement for strip steel surface defect,image segmentation for strip steel surface defect and image classification for strip steel surface defect,respectively.The main research contents of this dissertation include the following aspects:(1)An image enhancement method for strip steel surface defect based on IALO is proposed.Three strategies are proposed to solve the problems of poor exploration capability and low searching accuracy of ALO.Firstly,a Lagrange inertia weight is designed,which can better balance the relationship between the algorithm's exploration searching and exploitation searching.Secondly,a random perturbation invasive weed strategy is proposed,and combined with the ALO through the method of segmentation,which not only avoids excessive increasing the amount of computation,but also improves the searching performance of the algorithm.Finally,an adaptive local search strategy is proposed,which accelerates the convergence rate of ALO.Through the combination of IALO and local/global enhancement model(LGE),the image enhancement task for strip steel surface defects is completed.The experimental results show that the proposed image enhancement method for strip steel surface defect of this dissertation can better improve image contrast and highlight details of defect when comparing with common image enhancement methods,and the IALO has obvious advantages in solving the image enhancement problem for strip steel surface defect when comparing with the similar algorithms.(2)An image segmentation method for strip steel surface defect based on ABA is proposed.Four strategies are proposed to solve the problems of the poor adaptability and low searching accuracy.Firstly,an intelligent inertial weight is designed,which can intelligently adjust the flight speed of bat according to the number of iterations and fitness values.Secondly,a Beta distribution strategy is proposed,which improves the algorithm by adaptively adjusting the bat searching frequency.Thirdly,the local search strategy is improved,and only poor bats with poor fitness values can enter the local search with a certain probability,and the searching performance of the algorithm is further enhanced.Finally,an elite crossover strategy is proposed,and the cross operation is implemented on the optimal solution of the current iteration and the global optimal solution,which improves the exploitation capability of the algorithm.Through the combination of ABA and the maximum between-class variance method(Otsu),the image segmentation task for strip steel surface defects is completed.The experimental results show that proposed image segmentation method for strip steel surface defect of this dissertation can better segment defect when comparing with common image segmentation methods,and the ABA has better searching performance in solving the image segmentation problem for strip steel surface defect when comparing with the similar algorithms.(3)An image classification method for strip steel surface defect based on PCA-GOA is proposed.Three strategies are proposed to solve the problems of GOA in terms of poor exploration capability and unreasonable processing methods of transboundary grasshoppers.Firstly,an improved adaptive parameter is proposed to enable the grasshoppers with poor fitness values can move with a longer distance,and a parameter compensation mechanism is used to adjust the exploration capability of the algorithm with a flexible way.Secondly,the principal component analysis method(PCA)is used to generate unrelated grasshopper individuals to replace the individuals with low qualities,therefore,the searching performance of the algorithm is improved.Finally,an exponential boundary mutation strategy is designed,which can gradually place the transboundary grasshoppers near the boundary,which improves the processing level of transboundary grasshoppers.Through the combination of PCA-GOA and support vector machine(SVM),the image classification task for strip surface defects is completed.The experimental results show that that proposed image classification method for strip steel surface defect of this dissertation has a higher classification accuracy when comparing with common image classification methods,and the PCA-GOA has significant advantages in solving the problem of image classification for strip steel surface defect when comparing with the similar algorithms.(4)To test the effect of the proposed image processing methods for strip steel surface defect of this dissertation in the real environment,a performance test system of image processing methods for strip steel surface defect is built.The image enhancement method for strip surface defect based on IALO,the image segmentation method for strip surface defect based on ABA,and the image classification method for strip surface defect based on PCA-GOA are tested.The test results show that the proposed image processing methods for strip steel surface defect of this dissertation has significant advantages over the other comparative methods in the real environment.Finally,a strip steel surface defect image processing system based on swarm intelligence optimization algorithm is developed according to the research results of this dissertation.