The Classification Of Cervical Cell Pathological Image Recognition Method Research

Using computer image processing, pattern recognition technology on cervical cells image classification on the cervical cells is of great significance in the diagnosis of cervical precancerous lesions. On the basis of the existing research, combined with the basic knowledge of cervical cell pathology, this paper analyzes and implements the cell segmentation, the cell feature extraction and classification cell image. The main research contents involved : Preprocessing of cervical cell image segmentation, Preprocessing of cervical cell image segmentation, single-celled precise segmentation; To extract the characteristic parameters for the objective and effectively used for classification and recognition of the characteristic parameters(shape, color, texture), The research work of this paper is divided into two parts:The segmentation process of cervical cells. Firstly pre-cell image segmentation, and the foreground of the cell was determined from the intact cervical LCT images. Image segmentation is divided into three regions: single cell, cell group, the impurity. Because cervical cell images are huge picture resolution of up to tens of thousands of pixels, so using block binding Otsu segmentation algorithm to speed up the speed. The overlapping is inevitable problems in the process of production. Secondly, the segmentation of overlapping cell images will affect the accuracy of the parameter extraction of cell morphological characteristics by separating the cell group into independent cell segmentation results. Need to determine whether the cell population is divided can be divided into cell populations with separation points based on curvature method for the detection of overlapping cells through separation based on curve fitting method, indivisible is discarded. The third step is the accurate segmentation of single cell, the cell is a cell precise segmentation feature extraction accurate and effective premise, this paper uses a segmentation method based on Snake model,adaptive GVF experiments show that the method can achieve accurate segmentation, and satisfactory results.BP neural network was used to classify the cervical cells, and the standard BP neural network was used to consider its own inherent defects, According to the authority of the cervical cell pathological diagnosis standard(TBS), screened 15 alternative characteristic parameters, Then to them were principal component analysis, obtained a necessary and sufficiently small feature set, through the experimental comparison shows that the feature set can greatly shorten the time of training, so as to design a stable and reliable model. Finally, in the process of training to introduce the adaptive vector and momentum factor to adjust theweights between the layers to speed up the network convergence speed. In this paper, we study the process of using computer technology combined with a pathologist of On the basis of practical experience in,and use of image processing technology for image processing, the application of artificial neural network to classify the cells, with research and clinical diagnosis of certain medicine practical significance and broad application prospects.