Automatic Recognition Of Urinary Sediment Image Based On Image Processing

Automatic urine sediment detector is the one of main means to achieve automatic detection of urine sediment in hospital,the recognition rate and identification of the speed in urine sediment for automatic detection is very important.This paper researches the algorithm of automatic recognition of urinary sediment images,introduced the achievements at home and abroad widely,and combined with the latest development of the technology and the urinary sediment image features from the image filtering and enhancement,feature extraction and selection,image segmentation,image recognition and other aspects of urinary sediment image recognition.In the pre-processing,the Gauss filter is used for preprocess the image,then the gamma transform is used for the image enhancement,which makes the image smoothing and contrast become large.In the steps of image segmentation,At first,the OSTU method is used for image pre-segmentation which segmentation of foreground and background,Second,the model of anomaly detection is established that image segmentation will be carried out to remove the noise.compared with the traditional method,this model can get the quantitative analysis results of image segmentation and the segmentation result is more convincing.In the feature extraction,the morphological features of the cells are extracted.Compared with other features,the morphological features are more physical,and the PCA(Principal Component Analysis)algorithm is used to select the features.In the cell classification,this paper uses machine learning classification algorithm,through the comparison of several machine learning algorithms and combined with the background of urinary sediment image recognition,the classification method of SVM(Support Vector Machine)to classify cells.The MATLAB platform and the LIBSVM public library is used to test image with the algorithm in this paper,and based on the statistical test,the result show that the algorithm is not only accurate to filter impurities in urinary sediment images,and can distinguish more efficient in red cells and white cells,the final classification accuracy rate reaches 95%.