Study On Segmentation Algorithm Of Plant Roots CT Image Sequences Based On Prior Structure

In the study of geological structure,the factors such as root distribution,gravel parameters and pore structure in soil layer are important indexes reflecting the quality of soil layer.With the development of nondestructive testing technology,computer tomography(CT)has been widely used in the observation of soil structure and root morphology due to its non-destructive characteristics.Because of the detector response and the radiation intensity distribution of the radiation is not uniform and other factors,making CT image sequence reconstruction results of each layer slice’s gray level inconsistent in the vertical direction during the three-dimensional CT imaging process.The soil slice image itself has the characteristics of vague boundary and complex structure.These problems bring some difficulty for image object extraction.Based on the continuity and regularity in three-dimensional space,the structural information of the object region is used as the prior information to perform adaptive segmentation,so that to achieve the automatic segmentation of the CT image sequence.First,based on the continuity of the target structure,we came up with an adaptive segmentation algorithm of CT image sequence.Through the method of recursive iteration,the change of the target area is taken as the prior information,and the appropriate threshold is selected automatically for the image.And the feasibility of the method is verified by simulation experiment.Then to improve the under-segmentation in adaptive algorithm,we came up the random walk algorithm base on the previous structure continuity.Change the image gray-scale according to the threshold obtained by the adaptive algorithm,and mark the segmentation target by the super-pixel with entropy rate segmentation(ERS)method.Afterthen,according to the structural continuity the method of automatically dividing the seed points is proposed.Use the obtained effective maker points as the seed points of the random walk algorithm to segment the image.This algorithm effectively solves the problem of under-segmentation and improves the accuracy of 3D visualization.Finally,the effectiveness of the algorithm is verified by experiments.