ACF-based urine sediment detection method

Abstract

The invention discloses an ACF-based urine sediment detection method. Different detectors are trained for different urine sediment visible components. The method includes the following steps: (1) training: performing path calculation on a training sample, including LUV, gradient amplitude and the gradient direction histogram in six directions, then pooling ten paths, vectorizing the ten paths to obtain a feature vector, and training classifiers by using a soft cascaded adaboost algorithm, wherein each weak classifier is a decision making tree with the depth of 2; and (2) testing: extracting ten path features from a sampling window by using a sliding window detection technology, vectorizing the ten path features to form feature vectors, improving the detection speed by using a fast featurepyramid, performing test by using a trained model, and judging whether the sample belongs to the urine sediment visible components or not. The method has the advantages of good utilization of the feature information of the urine sediment visible components, effectiveness in reducing the influences of noises, high accuracy, fast calculation speed and great practical values.

Description

technical field [0001] The invention relates to the technical field of detection, in particular to an ACF-based urine sediment detection method. Background technique [0002] Urine sediment detection is one of the routine testing items in hospitals, and it plays an important role in the diagnosis and differentiation of kidney diseases, urinary system diseases and infectious diseases. For example, an increase in erythrocytes will indicate urinary tract bleeding, and the location of bleeding can be determined by further examination of the shape of red blood cells; an increase in leukocytes will indicate urinary system infection; increased casts indicate glomerulonephritis, renal tubular and renal dysfunction, etc. Therefore, urine sediment examination is of great significance. [0003] Urinary sediment inspection refers to the examination of the sediment (formed components in urine) after centrifugation with a microscope, which is to detect and count the formed components in ...

AI Technical Summary

Problems solved by technology

However, for some detection samples with a lot of noise and a lot of interference, traditional segmentation methods cannot effectively segment the background and target areas very accurately.

There are some researches on urine sediment detection in China, (see reference: Ding Ran. Image recognition and classification algorithm research on formed components of urine sediment: [Master's Thesis]. Jilin University, 2006.) Use prewiitte operator for edge detection, using two The dimensional entropy method thresholds the edge image, uses morphology to process the segmentation results, uses ellipse fitting to extract features, uses spectral analysis to extract texture features, and uses BP neural network for classification. More, cannot extract accurate morphological features, and the recognition rate of red blood cells and white blood cells is not high

You Yingrong et al. (see references: You Yingrong, Fan Yingle, Pang Quan. Adhesive cell segmentation method based on distance transform [J]. Computer Engineering and Application, 2005, 41(20): 206-208.) proposed a method for adherent cells A method based on distance transformation, in the case of not serious cell adhesion, this method can realize the segmentation of adhesion cells, but it cannot guarantee the original shape of cells

The step of image segmentation is very critical, and often image segmentation cannot achieve good results in the case of complex images

Images