The Research On Segmentation Method Of Ochotona Curzoniae Image In Natural Scene

Ochotona curzoniae image segmentation is the basis for target detection,tracking and behavior analysis of Ochotona curzoniae.The Ochotona curzoniae image on nature scene has the characteristics of complex background,intensity inhomogeneity,weak edges,low contrast,and diversity and graduality of target colors,which make the segmentation of Ochotona curzoniae images more difficult.In order to resolve the problem that the internal fitting value of Chan?Vese model can not effectively represent the integral information of the target when segmenting Ochotona curzoniae images with diversity and graduality of target colors and complex background,a new approach of the Chan?Vese model in combination with fuzzy Cmeans clustering is proposed in the present paper.The proposed model utilises fuzzy C-means clustering to cluster the pixels inside the evolution curve of the Chan?Vese model,classifying the pixels into a certain color cluster with a certain probability to describe the image color gradual characteristics.By fuzzy C-means clustering,several cluster centers can be obtained,and the values of cluster centers can be used to replace internal fitting values of the Chan?Vese model.In this way,the problem that the Chan?Vese model cannot segment images with diversity and graduality of target colors is overcome.Furthermore,the global Heaviside function is replaced by the local Heaviside function to suppress the influence of the background on image segmentation.The experimental results of Ochotona curzoniae images segmentation demonstrate that compared with the traditional and the latest level set image segmentation models,the proposed model has a higher Dice similarity coefficient,Jaccard Similarity,and segmentation accuracy,which shows that the proposed model can more accurately locate the target and has a better segmentation effect.The Chan?Vese model uses the global information of the image to construct the fitting values,so it is poor for the image segmentation with intensity inhomogeneity.In order to resolve the problem,this paper proposes a hybrid active contour image segmentation model which combines the local and global information of the image.Firstly,the local energy term is constructed by using the image intensity mean of the local region inside and outside the evolution curve to capture the intensity inhomogeneity of the image;secondly,In order to avoid the evolution curve of the active contour model based on local information falling into local optimum in the evolution process,the global energy term is constructed by using the intensity mean inside and outside the evolution curve to drive the evolution curve to evolve to the target edge;finally,the weight coefficient is constructed by the gray level of the local and global regions of the image to adaptively adjust the relationship between the local energy term and the global energy term,so as to make the evolution curve more stable,so the model can adaptively adjust the evolution of the curve with the change of the target region.The segmentation experiments on natural images and brain tumor images show that compared with the traditional and some latest active contour models,the proposed hybrid active contour image segmentation model has higher segmentation accuracy and is insensitive to the initial contour.