Research And Classified Recognition Of Acute Leukemia Cell Based On Digital Image Analysis

In this thesis,the original cell images are preprocessed and some samples are obtained.The feature parameters of morphology are extracted from the images of cell samples.The images of cells samples are recognized and classified by BP Network.On the part of segmentation,we adopt the watershed segmentation method based on the Wavelet transform,settling the problem of over-segmentation and segmentation error very well.After the division transforms into two value pictures, we use an eight chain code algorithm to get to include the perimeter,the area of resembles,the shape,corn compared to cytoplasm and so on.Several main morphology parameters is carried on the survey.Finally,the obtained the massive data samples are taken into the nerve network for training.In this thesis the nerve network model mainly bases on the BP reverse propagated error algorithm.After obtaining the massive data sample,we begin to train the network.After the error is smaller than the rating,it can be carried on the test in the application data sample to the nerve network.By the massive experimental contrast results,we make the conclusion that the BP reverse propagated error algorithm has a quite good classification ability to acute leukemia cell.