Analysis And Research On Erythrocyte Morphology Based On Clustering Algorithm

With the development of the economy and continuous improvement of people's quality of life and cultural level,people increasingly pay more attention to their own and family health care.Human health is closely related to red blood cell morphology and quantity;many diagnoses of diseases based on blood cell morphology and quantity.In this paper,according to the requirements of red blood cell automatic identification system,the use of image processing and pattern recognition technology,make research to the key technology,with the experimental demonstration which conclude a better solution to the problem of red blood cell classification and identification.This paper studied and experimented aspects of the following items that are based on the research of red blood cell microscopic image automatic recognition system,the red blood cell semantic model,the red blood cell image segmentation algorithm,the red blood cell feature extraction and selection,the fuzzy clustering classification recognition.It is mainly to complete the following work:1?In the aspect of image segmentation,after analyzing the characteristics of red blood cell image and partial segmentation algorithm,a red blood cell segmentation algorithm based on graph cut is proposed.The algorithm is to extract the target from a complex background by iterative way.The influence of the problem of the irrelevant region and the internal cavity on the segmentation and the problem of the separation of the multi-adhesive cells are solved,which reduces the complexity of the algorithm and improves the segmentation speed and segmentation accuracy.2?In the shape feature extraction,the basic morphology of all kinds of erythrocytes was studied,and then semantic model classifier was established.According to the semantic model,the mathematical expression method of the model feature and two new shape features are established to improve other shape features which provide a basis to further identify a variety of shaped erythrocytes and improve the operating efficiency and classification accuracy.3?In the texture feature extraction,the texture features of red blood cells are described by covariance matrix analysis.After comparing the experimental data,the following items including the second order moments,contrast,correlation,entropy,inverse moment,and entropy and mean difference,difference entropy,variance and nine texture features are chosen.4?The basic principle of fuzzy clustering and the FCM clustering algorithm are studied.The speed and accuracy of the algorithm are improved under researches of The shortcomings of the algorithm.FCM algorithm was used to cluster 11 erythrocytes,and the clustering results were analyzed.Experiments show that the improved clustering algorithm is used to realize the efficient operation of the code and to improve the efficiency of the image data processing to meet the engineering requirements of the real-time image processing,which completed the classification of the red blood cells and get the better classification effect.