Study On Blood Cell Morphological Examination And Recognition Technology

Blood cells contain red blood cells,white blood cells and platelets.The morphological examination of white blood cells provides a strong basis for clinical diagnosis and has broad application prospects.At present,it is a common situation to observe the blood cell slide by artificial microscopy,but it is boring and inefficient to classify the white blood cells artificially.The automation and intellectualization of white blood cell detection through image processing and pattern recognition technology is the current development trend.Based on motion control platform,color CCD camera,optical microscope and high-performance computer,this paper builds an automatic blood cell detection system.High-quality white blood cell microscopic images are obtained by using efficient image acquisition scheme and automatic focusing technology.The white blood cell image online recognition is realized by image segmentation and classification algorithm.In image acquisition,two-dimensional discrete wavelet transform is used to locate the Monolayer area under low power microscopy,and spot detection is used to locate white blood cells,then image acquisition is performed under oil microscopy.For image segmentation algorithm,according to the characteristics of computer micro-vision images,we select HSI color model which is in line with human eye observation to exfoliate white blood cells from background image accurately by screening image color model.By comparing and studying various image segmentation algorithms,region growing algorithm shows good segmentation performance on H channel of HSI model,which meets the next segmentation requirements.In the extraction of nucleus and cytoplasm,on the basis of background exfoliation of leukocytes,the white blood cell image of channel B of RGB model was extracted by Otsu method.In the aspect of classification and recognition,the feature library of morphology,color and texture of white blood cells was established.Based on the principle of small quantity,reliable,independent and distinguishable feature extraction,the parameters for characterizing the morphological characteristics of white blood cells were screened out.Support vector machine(SVM)and artificial neural network(ANN)are used to train the characteristic parameters.After comparing the classification accuracy of three kernel functions of SVM with that of ANN,a morphological classifier based on BP neural network is designed to recognize white blood cells with high accuracy.Based on the above research,a leukocyte classification and recognition system of blood cell image based on computer micro-vision platform is developed,which realizes the functions of high-quality scanning,acquisition,localization,segmentation and classification and recognition of leukocyte image,and replaces the artificial microscope in clinic.It improves efficiency,reduces labor intensity and meets the needs of morphology of leukocyte in clinical medical micro-vision automatic detection.