The Blood Cell's Recognition And Count Based On Improved Division Algorithm

In the medical microscopic image recognition, image segmentation is a critical technology and a classic problem. In practice, for different features of the image using the corresponding segmentation model and algorithm, which restricte the medical image processing technology, to some extent. With the development of computer technology, the phenomenon that digital image processing is applied to medical research has become a heated topic, automatic analysis and recognition of blood image is one of the representative topics that computer is used by medical microscopic image processing.With the introduction of artificial intelligence and automatic identification in the blood cell analyzer, the introduction of digital image processing used by blood cell analyzer becomes a trend.Specific to the characteristics of blood cells overlap, we sum up the threshold segmentation, edge detection segmentation, mathematical morphology segmentation,chain code segmentation, bring up an improved image segmentation algorithm and achieve cell image segmentation and count successfully which based on Delphi programming.The paper deals with the theory research of segmentation algorithm, implement the comparison of effects picture by programming and analyse the positioning accuracy and noise level. Specific to the characteristics of blood cells overlap, it proposes the improved segmentation algorithm which is on the basis of threshold segmentation and edge detection segmentation with the complement of mathematical morphology knowledge.The chain code algorithm which realizes the segmentation and statistics of cell is on basis of threshold segmentation, utilize technology of chain code table and segment table. According to the cell image features cell is divided into noise, single cell and adhesion cells by comparing size,target shape factor and the circular degree; it is applied to segmenting adhesion cell images using curvature segmentation method of chain code This article compares target form factor and roundness, removes impurities and segments cell image; compared again to prevent over-segmentation phenomena; on the cumulative count of the cells after division.We accomplish the training and identification of blood cells by using BP neural network algorithm and the segmentation and counting system of cell by programming.