| Image segmentation is the basic technology of image engineering and the key process of image analysis and understanding.However,the uncertainty,ambiguity and complexity contained in images often cause traditional algorithms to fail to obtain good segmentation results,which affects subsequent analysis of results.The introduction of fuzzy clustering theory is really good at expressing such characteristics of images.Therefore,it has been applied by many scholars in the research of image segmentation,which proves that the application of fuzzy clustering theory can get more ideal segmentation results,which is superior to traditional algorithms.Fuzzy C-Means(FCM)clustering algorithm has been recognized by many scholars in the image field.The mathematical theory of the algorithm is perfect and it has a wide range of applications.The FCM algorithm continuously iterates the membership matrix and clustering center to divide which category all pixels in the image belong to,and obtains the minimum value of the objective function,which reflects the fuzzy characteristics of the image and avoids some other problems of segmentation algorithms.However,the algorithm is not perfect,and it also has some defects.For example,when initializing the parameters,the deviation is too easy to fall into the local minimum;it simply calculates the information of each pixel,and does not calculate the neighborhood information of the pixel,so when the image is complicated or it contains noise,it will affect the results;the number of iterations increases,the amount of calculation is large,and it takes too much time.This thesis mainly studies the fuzzy clustering image segmentation algorithm.In view of the development status of image segmentation and the problems exposed by the fuzzy clustering algorithm in the application of image segmentation,the following contents have been studied:(1)The research background and significance are elaborated,and the contents of commonly used segmentation algorithms,such as thresholds,region growth,edge detection,and methods of fusing other theories are specifically discussed.The basic knowledge of fuzzy clustering is introduced in detail,and the current research status and problems of fuzzy clustering algorithms are analyzed.Then,we focus on the standard FCM clustering algorithm,the bacterial foraging algorithm and its initialization parameters,and summarize the advantages and disadvantages of the two algorithms in image segmentation.(2)The initial clustering center of fuzzy C-means clustering algorithm is set artificially.If its setting deviates too much from the global optimal clustering center,or if the algorithm has problems during the iteration process,it will cause the FCM algorithm to fall into the local optimal value.Aiming at the above problems,this thesis proposes an adaptive imagesegmentation method based on FCM clustering for adaptive bacterial foraging.With the help of bacterial foraging(Bacterial foraging algorithm,BFO)algorithm,it has the advantages of strong parallel searching and easy to get rid of local optimization,which is very suitable for solving more complicated optimization problems.However,when the BFO algorithm is used to solve different problems,the parameters are not adaptive,which may cause the convergence speed to drop and the accuracy of the solution to be inaccurate.This thesis uses Tent mapping to initialize the bacterial population,and introduces adaptive dimension learning to improve the bacterial population to find the best foraging position.This avoids the BFO algorithm increasing the search time by changing all dimensional values of the bacteria,which leads to the convergence speed of the algorithm and the accuracy of the settlement is affected.Also,the FCM algorithm is combined to accelerate the convergence speed and avoid the algorithm from falling into local extreme values.(3)For color image segmentation,this thesis compares some features of RGB and HSI color spaces,and calculates the RGB space into HSI space through calculation.The color image segmentation is performed by using the proposed FCM clustering optimization algorithm of adaptive bacteria foraging.At the same time,because the applicable range of Euclidean distance in fuzzy C-means clustering algorithm is too limited,this thesis uses feature distance.Simulation experiments prove that this algorithm can also achieve ideal segmentation results for color image segmentation. |
