Segmentation And Recognition Of Urinary Sediment Image Based On Bayesian Network

With the development of the economy,people's living standards have been improving dramatically.People are also paying more attention to their own health.Human's urine composition can effectively reflect the people's health status,among which the type and quantity of the Urinary sediment can effectively reflect the health of the kidney.In this paper,based on the segmentation and identification requirements of Urinary sediment image,image processing techniques based on Bayesian networks are used.The key steps were studied and verified by experiments,which solved the problem of segmentation and classification of Urinary sediment image.Bayesian networks differ from other networks in that their knowledge forms are full of uncertainty.Probability is expressed in terms of diversity,and the knowledge gathered together to form new learning characteristics.Due to this nature,Bayesian networks are more suitable for solving medical image problems.This article mainly studies the segmentation and identification of Urinary sediment image.Methods for image preprocessing and segmentation algorithm of the Urinary sediment image,feature extraction and dimensionality reduction of Urinary sediment image,and identification and classification algorithm of Urinary sediment image have been studied and tested.The following aspects are included.(1)In image preprocessing,a set of combined preprocessing methods is proposed.This method first performs binarization.When the threshold is selected,the maximum grayscale value is added to the minimum grayscale value divided by 2,and then a random function is used to take a random number between zero and this value.This random number is used as the threshold for binarization;next,a pair of pixel structures are used for the etching process;then,the size of the pixel is used to delete some small objects;finally,the image is expanded three times using the 2x2 structure.This method reduces image noise and improves segmentation accuracy.(2)For urine sediment images,there are various problems such as uneven gray scale and large noise.This paper proposes that image segmentation is regarded as the maximum a posteriori probability estimation,and constructs the Bayesian probability model by constructing the energy function to solve the global minimum value of the energy equation.According to the segmentation result of the image,according to the 17 features after dimensionality reduction,a urine sediment image recognition algorithm based on Bayesian network was proposed.The extracted features were used to construct the conditional probability table.According to the sample data of each attribute,the condition of the feature was obtained.The probability table,thus constructing a Bayesian network.Bayesian network is used to realize the recognition of urine sediment images.The algorithm has high recognition accuracy,strong robustness and fast recognition speed,and has strong engineering application advantages.