The Key Technique Studies On Automatic Recognition Of Red Blood Cell Image

The count and recognition of red blood cells plays an important role in modern clinical practice. At the same time, it is the key foundation for diagnosing kinds of blood diseases and other pertinent diseases. With the digital image processing and analysis has been widely applied in many medical areas, advanced image processing and pattern recognition technology, which is used in the sum and sort counting of blood corpuscles, is one of the important methods in medical-aided diagnosis. The technology effectively reduced subjective influence, partly replacing the labor works and improving the disadvantages of blood corpuscle analyzing apparatuses used currently in hospital, for example it can't observe the shape of the cells, can't save the swatch of the test, and the machine is very expensive, and so on.The study use the technology of image processing and pattern recognition to classify the red blood cells in different shapes, in order to enhance the correctness and exactness; Bring forward a set of basic algorithms including pretreatment, feature extraction and classification; Most work has been focused on the image segmentation and feature extraction. Satisfying results have been reached by the selected feature parameters acting on erythrocytes classification.1. Study classical image segmentation method such as threshold segmentation, edge detection, watershed and etc, design a new algorithm for automatically separating overlap cells images. The algorithm is simple and effective, it can obtain every cell's original shape preferably.2. According to the eyeballing experience of erythrocytes classification, combining with the former work in feature extraction, quantification described morphology parameters, such as shape and size of cell area are extracted. Then extract texture feature and improve the expressions of round-grade on the basis of experiments, in order to improve classification accuracy. Finally form an effective feature subset. 3. Hierarchical support vector machine classifier is constructed for 12 species recognition. With a small quantity of samples, our proposed methods achieve a decent performance on erythrocytes classification.The subject is an application research in the image processing and recognition field of medicine. The results of this study have both theoretical and practical values.