| The classified recognition of white blood cell is an important content of generally blood inspection. In order to alleviate the workload of inspector, enhance the definition of inspection, satisfy medical research work, in this thesis, we analysis the structure and capability of automatic recognition system of leukocyte and an applied system structure using a PC, a microscope and a color CCD camera is proposed based on other researcher's study. The core of this system is consisted of data capture module, pretreatment module, character distilling module and pattern recognition module. These modules play a very important role in processing of classified recognition. In addition, we also design a data storage module to save the result.In data capture module, in order to enhance the quality of premier data, we formulated criterion for preparation of blood smear for microscopic examination. We compress data for alleviating calculation and store after data capturing.In image segmentation module, we analysis classical segmentation method and evaluate the effect. We also present a new segmentationmethod------ultra-blue .This method based on characteristic of that theproportion of blue is biggest in stain leukocyte nuclear.In cell counting, we present a new labeling arithmetic. Thisarithmetic is performing as recursion. It can label all connected regions by one scan, so the complexity of arithmetic is reduced.In character distilling module, we set up a character vectors which is consist of cell nuclear area, cytoplasmic area, cell perimeter, cell shape and so on base on experience of clinic inspector.In pattern recognition module, we select fuzzy K-neighbor method to design sorter, construct ruling tree base on experience of clinic inspector. At last, we use a lot of samplers to test the validity of automatic recognition system.This research also do plenty of work on software design, we adopt orient-object method to design system and complete all coding in VC++6.0. |
PDF Full Text Request