Image Segmentation Of Adhesive Circular-like Objects Based On Concave Point And Centroid Detection

The circular-like object images refer to the images in which the shape and contour of the objects are circular or close to circular.For example,images containing circular-like object such as cell,colony and nanoparticle can be called circular-like object images.In biological,medical and other fields,object clumps are often observed in the captured images,which brings great difficulties to analyze the image content.Therefore,this paper studies this problem,aiming to use the method of image processing to segment the adhesion circular objects existing in such images more effectively and accurately.The main contributions of this paper are as follows: First,this paper summarizes the development trend and research status in the field of adhesion image segmentation,and introduces two related algorithms.One is based on concave point detection and ellipse fitting,and the other is based on concave point detection and concave point matching.Combined with the results of the experiment,this paper summarizes the problems and shortcomings of these algorithms.Secondly,in order to obtain the feature points(concave point and centroid point)needed by this algorithm to segment the clump,this paper proposes a more accurate concave point detection algorithm and an improved centroid point extraction algorithm.In order to solve the inaccuracy of concave point detection in existing algorithms,the concave point detection algorithm in this paper can detect the concave point more accurately through multiple criteria.At the same time,in order to solve the problem that the original Fast Radial Symmetry(FRS)transform needs to set the priori radius range manually when extracting the centroid point,a modified method is proposed to extract the centroid points adaptively.Thirdly,a simple but effective structure-aware clump splitting algorithm is proposed in this paper by using the extracted feature points.Existing splitting algorithms for clump splitting based on concave point detection often match the contour segments or concave points,and then complete the clump splitting by ellipse fitting or constructing the split line.However,it is difficult to complete the correct matching or fitting for the more complicated clump only by using the concave point information,so that there are more over-segmentation and under-segmentation in the segmentation result.In this paper,a new approach of the centroid point assisted concave point matching is proposed.The possible types of clump are grouped according to the number of concave points and centroid points detected in each clump and the mathematical relationship between them.The possible types of clump can be grouped into four types: simple structure,tandem structure,cluster structure and tandem-cluster structure.According to the characteristics of each type of clump,the corresponding method is proposed to complete the clump splitting in circular-like object images.Finally,a novel post-processing technique is proposed to deal with the cases when the clumps have incomplete clump structures or the concave points are missed due to poor image quality.The algorithm of this paper uses Matlab programming language to implement the algorithm and related experiments in Windows 7 environment,and it is compared with four existing typical adhesion segmentation algorithms with multiple sets of experimental data.The results show that the proposed algorithm in dealing with the clump of the subjective and objective experimental results are better than the comparison algorithms.Therefore,the accuracy and robustness of the proposed algorithm are higher.Moreover,based on the algorithm of this paper,a set of colony image segmentation software is developed.The software is developed in C++ programming language under Windows 7 environment.The interactive interface is designed based on MFC,and the Open CV library and cv Blob library are used to complete the development and implementation of related image processing algorithms.The software mainly realized the colony image segmentation,statistical data and other functions.