Research On Image Segmentation Methods Based On Bionic Optimization

As a kind of multimedia data containing a large amount of information,image plays an increasingly important role in people's life and work.Image segmentation is a crucial pre-processing step for image recognition and computer vision.It is also an effective way to realize image understanding.As an important part of image processing,more and more attention has been paid to it.Image segmentation has been widely used in computer vision,face recognition,product detection,industrial automation,intelligent transportation,text recognition,alien detection,aviation and aerospace technology,remote sensing satellite image processing,biological and medical engineering,sports and agriculture and other fields.In many engineering applications,it is difficult to obtain a large number of image samples due to the complexity and diversity of practical problems and the limitation of various factors.Gray image is the most basic and most widely used of all kinds of images,with the advantages of simple representation,small amount of data,convenient coding and transmission,etc.As the basis of building complex images,gray image is suitable for most image processing occasions,and its segmentation method is also widely used in a variety of disciplines and engineering applications.Therefore,small sample gray image segmentation technology with high segmentation accuracy and computing efficiency is of great significance and engineering value.Image segmentation technology involves many interdisciplinary disciplines such as cognitive science and computer vision,and is also widely used in engineering practice.The segmentation method based on feature clustering and threshold is widely used and improved in the field of image segmentation because of its simple structure,strong applicability and high segmentation accuracy.However,the traditional gray image segmentation methods have low segmentation precision or poor computational efficiency.Bionic optimization algorithm can calculate the complex nonlinear multidimensional data space quickly and effectively,and the image segmentation problem can be equivalent to the problem of seeking the optimal segmentation parameter in the complex parameter space.In this paper,the image segmentation techniques based on feature space clustering and optimal threshold selection are studied,and Grey Wolf Optimizer algorithm with better computational stability and processing speed is introduced to improve the algorithm performance and image segmentation accuracy.The main research work and innovation points are as follows:1.In view of the uncertainty of the initial clustering center in the segmentation of images by the fuzzy c-means algorithm,the number of clustering categories needs to be set artificially,and it is easy to fall into the problem of local optimization in the iterative process.In addition,the search population simplification of the Grey Wolf Optimizer algorithm is easy to fall into the problem of local optimization and premature convergence,an image segmentation method based on differential evolution Grey Wolf Optimizer algorithm and fuzzy c-means is proposed.Since the estimation of clustering centers and their number in FCM algorithm can be regarded as the search process of finding suitable values in the gray range,Grey Wolf Optimizer algorithm based on differential evolution is used to find clustering centers.By introducing the idea of differential evolution to improve the diversity and mutation ability of wolves,the algorithm can avoid falling into the local minimum value,and the segmentation accuracy can be further improved.2.In view of the problems of traditional fuzzy c-means algorithm,which only uses the membership degree information of pixels,does not make full use of the spatial information around pixels,is sensitive to noise and uneven gray value,has low segmentation accuracy and poor robustness,etc.,a fuzzy c-means image segmentation method based on a parallel LGWO and local information is proposed.The Levy flight strategy is introduced to improve the computational efficiency and search performance of the algorithm,and the design of parallel computing program can greatly improve the computational efficiency of the algorithm.And the noise image is processed with adaptive grayscale weighting based on the neighborhood information of the image.The proposed algorithm can effectively suppress the influence of the noise and improve the segmentation accuracy.3.When the noise pollution of the image is serious,the neighborhood information of the image pixel may also be polluted,resulting in the reduction of the segmentation accuracy of the fuzzy clustering algorithm combined with the local spatial information of the image,which cannot meet the requirements of high-precision segmentation.Therefore,a fuzzy c-means image segmentation method based on an improved parallel LGWO and global information is proposed.Firstly,the parallel LGWO algorithm improved by the new convergence factor and dynamic weight strategy is used to roughly cluster the images to get the initial clustering center.These improvements can make the parallel LGWO algorithm have higher search accuracy.Secondly,the neighborhood information and non-neighborhood information around the pixel are added to the objective function as spatial information,the information entropy is used to adjust the weight between the pixel information and non-neighborhood spatial information,and the improved distance measure is used to replace the traditional Euclidean distance,so as to further improve the segmentation effect of the algorithm.4.Because the choice of threshold directly affects the segmentation accuracy of threshold segmentation technology,traditional threshold segmentation methods are mostly obtained through traversal,affecting the computational efficiency of the algorithm.In this paper,a multilevel threshold image segmentation method based on improved Grey Wolf Optimizer algorithm is proposed.Tsallis gray entropy threshold technology is used to determine the optimal threshold,which has the characteristics of simple and easy to implement,high segmentation accuracy,and easy to expand from single threshold segmentation to multi-threshold segmentation.On the basis of the previous studies,the population initialization strategy of Opposition-Based learning,Levy flight strategy,adaptive boundary strategy and poor individual active gaussian mutation strategy are introduced to improve the Grey Wolf Optimizer algorithm.The improved algorithm can better balance the search and development ability of the algorithm,avoid the algorithm from falling into local optimality,and improve the convergence speed,global search ability and segmentation accuracy of the algorithm.The simulation results of each part further demonstrate the effectiveness of the proposed method.