Multi-objective Cuckoo Search Fuzzy Clustering For Image Segmentation Algorithm And Its Application

Traditional fuzzy clustering image segmentation methods,such as fuzzy C-means clustering algorithm,have difficulty in meeting the actual needs of the balance between noise effectiveness and detail retention due to poor robustness and vulnerability to noise.Researchers rely on the algorithm to add objective functions with local spatial information or non-local spatial information to control the parameter connection of multiple objective functions.The existence of this control parameter makes it difficult to accurately obtain the optimal solution and thus affect the effect of its image segmentation.Based on the multi-objective evolutionary algorithm with global search performance,the image segmentation of fuzzy clustering can be transformed into multi-objective optimization problem.Multiple objective functions can be optimized simultaneously without control parameters,and more image information can be used to solve complex pixel distribution,to improve the performance of image segmentation and to achieve a balance between complex image information and detail retention noise immunity.Therefore,this paper studies the image segmentation algorithm based on fuzzy clustering and combines with multi-objective Cuckoo Search algorithm based on decomposition(MOCS/D),a decomposed multi-objective Cuckoo Search fuzzy clustering for image segmentation algorithm(MOCS/D_FCM)is proposed and applied to the surface roughness measurement of grinding specimens based on machine vision.The main research contents of the thesis are as follows:(1)This paper reviews the domestic and international research status of image segmentation methods based on four aspects: edge,region,specific theory,clustering and the machine vision surface roughness measurement.The problem of image segmentation algorithm based on fuzzy clustering in dealing with images with strong noise level is obtained,and the research content and ideas of the thesis are determined.(2)For the fuzzy clustering improved algorithm combining local spatial neighborhood information and non-local spatial neighborhood information,the key parameters are connected to the objective function,which leads to the problem that the image segmentation effect is not ideal.A MOCS/D_FCM is proposed,and this clustering image segmentation algorithm is applied to the segmentation of synthetic images and natural images to verify the effectiveness of the proposed algorithm in performance evaluation.(3)Aiming at the problem that the current surface roughness measurement method based on machine vision generally fails to make full use of the fuzzy information of the imaging pixel on the workpiece surface,the algorithm is applied to the roughness detection of the grinding sample,and the roughness evaluation index is used,the maximum membership index C and fuzzy information index F;The virtual images of the 16 sets of surface roughness grinding blocks based on red and green blocks are segmented,the fuzzy membership matrix of the virtual image is obtained,and two roughness evaluation indexes are solved to establish a prediction model.The feasibility of applying the algorithm to surface roughness measurement is proved by the goodness of fit.(4)Based on LabVIEW and MATLAB platform,a machine vision-based surface roughness measurement software was developed using MOCS/D_FCM,and it was applied to the measurement of surface roughness of 8 different grinding samples,realizing automatic image acquisition and roughness detection function.