Pattern Recognition Research Of Microscope Wet Leucorrhea Image Based On CNN-SVM

The wet leucorrhea smear microscopic image of candida and white blood cell detection are the two important indicators of inflammation in gynaecology department diagnosis,because of high incidence of inflammation in gynaecology department clinical diagnosis and leucorrhea smear reading work relies too heavily on physicians’ naked eyes, which makes the diagnosis efficiency is extremely low; With the computer pattern recognition technology applied in biomedical, the medical microscopic image target detection identification becomes possible to be intelligent, because the wet leucorrhea microscopic image’s target and its background appear obscure, weak edge, single color gray features on the whole and the texture between target and its background is not obvious,which makes the wet leucorrhea microscopic image object intelligent detection very difficult, resulting that the related research is blank.the leucorrhea microscopic image represents a feature that is difficult to detect, at this point, this paper revolves around object detection of the leucorrhea microscopic image intelligent, studies the technologies of the white blood cells and candida images’ segmentation, feature extraction, classification and recognition under the leucorrhea microscopic, which has obtained the comparatively abundant research results. the specific work contents are as follows:1,the leucorrhea microscopic image ROI(Region Of Interest, ROI) form and the differences of background and ROI area are explored, the optimal target segmentation method is that the morphological opening operator was used to the first round of denoising,within the scope of the large radius, white blood cells is directly detected by the method of Hough circle detection, which gets the high recognition rate, and gives the best Hough circle detection radius on the basis of a large number of experiments.2, based on the form of candida appearing like egg or "8" and the image on the whole prospect presents the characteristics of particles under wet leucorrhea microscopic image,although the candida coverage is up to 100% under Hough circle detection within the scope of small radius,the candida and error detection of candida area is positioned, the candida learning pattern library was first established, and the best radius of Hough circle was given.3,Based on the previous work, according to CNN-SVM model,the convolution network was used to get the feature of the candida and noise area got by Hough circle detection within the scope of small radius, the SVM classifying algorithm was used to classify the feature to recognize the candida,the higher candida detection rate was obtained.4,Based on previous research, white blood cells and candida detection system was designed, the CNN-SVM classification model evaluation was given.The value of this research is that the intelligent detection of related disease of gynaecology department caused by candida is realized,The results will be directly applied to the wet leucorrhea microscopic image intelligent testing equipment’s software system,which has great practical application value.