CN104751167A - Method and device for classifying urine visible components - Google Patents

Abstract

The invention provides a method and a device for classifying urine visible components. The method comprises the following steps: using the difference between the average pixel value in a peripheral region of the visible components inside a visible component block and the average pixel value in a central region as characteristics in a characteristic set of a classifier; and using the classifier to classify the urine visible components. An embodiment of the invention aims at increasing the precision of classifying the urine visible components.

Description

Urine formed component classification method and device

technical field

The invention relates to biological detection, in particular to a method and device for classifying urine formed components.

Background technique

In a common urine sediment analysis technique, firstly, a microscope system is used to take an image of a urine sample. Then, the particle block in the urine sample image is segmented by edge detection technology. Impurity particle blocks are removed from these particle blocks, leaving blocks of formed elements (eg, red blood cells, white blood cells, crystals). Then, these formed components are classified, for example, into red blood cells, white blood cells, crystals, and the like.

The classification process often uses a classifier based on the trained model. For example, some features that are helpful for distinguishing various formed components, such as area, circularity, extension, gradient, etc., will be used to form a feature set to train a classifier, such as a neural network. The classifier is trained by using a large number of samples of formed partition blocks and measuring their characteristics (such as area, circularity, stretch, gradient, etc.). In this way, when a new formed block is input, the trained classifier will classify it according to the calculated features of the block.

Common classification techniques use shape features and some texture features to form feature sets. The features in the traditional feature set are not effective enough to distinguish the categories of different formed components, and the classification accuracy is not high enough.

Contents of the invention

One embodiment of the present invention aims at improving the accuracy of urine formed component classification.

According to one embodiment of the present invention, a method for classifying urine formed components is provided, comprising: using the difference between the average pixel value of the peripheral area of the formed component in the formed block and the average pixel value of the central area as a classifier The features in the feature set; the urine formed components are classified by the classifier.

In a specific implementation, the step of using the difference between the average pixel value of the peripheral region of the formed component and the average pixel value of the central region as a feature in the feature set of the classifier includes: removing the influence of ambient light and out-of-focus noise on the formed component Influence of the pixel value in ; calculate the difference between the average pixel value of the peripheral area and the average pixel value of the central area after removing the influence of ambient light and defocus noise; use the difference as a feature in the feature set.

In a specific implementation, the step of removing the influence of ambient light and defocus noise on the pixel value in the formed component includes: for the pixel in the formed component, calculating

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. color factor,

${\mathrm{<span\; class="notranslate">\; NT</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}$

max(T _red ), max(T _green ), max(T _blue ) represent the maximum value of the three-color coefficient of red, the maximum value of the green three-color coefficient, and the maximum value of the blue three-color coefficient of the pixel in the formed component, NT _green represents the normalized value of the green three-color coefficient of the pixel;

The step of calculating the difference between the average pixel value of the peripheral region and the average pixel value of the central region after removing the influence of ambient light and defocus noise includes: calculating the average value of NT green of each pixel in the peripheral region of the formed component and the NT _green value of the central region The difference between the average values of NT _green for each pixel.

In a specific implementation, the step of removing the influence of ambient light and defocus noise on the pixel value in the formed component includes: for the pixel in the formed component, calculating

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

T _gray =0.2989T _red +0.5870T _green +0.1140T _blue

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. Color coefficient, T _gray represents the gray coefficient of the pixel,

${\mathrm{<span\; class="notranslate">\; NT</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}<span\; class="notranslate">\; )</span>}$

max(T _gray ) indicates the maximum value of the gray coefficient of the pixel in the formed component, and NT _gray indicates the normalized value of the gray coefficient of the pixel;

The step of calculating the difference between the average pixel value of the peripheral area and the average pixel value of the central area after removing the influence of ambient light and defocus noise includes: calculating the average value of NT _gray of each pixel in the peripheral area of the formed component and the average value of NT gray in the central area. The difference between the average values of NT _gray for each pixel.

In a specific implementation, the method further includes: obtaining the gray level co-occurrence matrix of the formed component, and using one or more of the contrast, homogeneity, and energy of the gray level co-occurrence matrix of the formed component as the classification The features in the feature set of the device, where the contrast, homogeneity, and energy of the gray-level co-occurrence matrix are calculated as follows:

Among them, P(i, j) represents the probability that the values of gray level i and gray level j appear at the same time.

In a specific implementation, the method further includes: using the difference between the average brightness of the pixels in the foreground part of the formed sub-block and the average brightness of the pixels in the background part of the formed sub-block as a feature in the feature set of the classifier.

In a specific implementation, the method further includes using one or more of the seven Hu distances I ₁ , I ₂ , I ₃ , I ₄ , I ₅ , I ₆ , and I ₇ calculated according to the following process Features in the feature set for the classifier:

Convert the formed component into a grayscale image, f(x,y) represents the grayscale of the pixel (x,y) in the grayscale image, f(x,y)=0.2989I _red +0.5870I _green +0.1140I _blue , Among them, I _red , I _green , and I _blue represent the pixel values of the pixel (x, y) in the red layer, green layer, and blue layer respectively, and the (p+q) steps are defined as:

m _pq =∫∫x ^p y ^q f(x,y)dxdy p,q=0,1,2,…

(p+q) order center distance is defined as:

μ _pq =∫∫(xx ₀ ) ^p (yy ₀ ) ^q f(x,y)dxdy

in $x_0=\frac{m_{10}}{m_{00}},{\mathrm{the\; y}}_0=\frac{m_{01}}{m_{00}}$

The standard center distance is

in $r+\frac{p+q+2}{2}$

Thus, the seven Hu distances are defined as:

I ₁ =y ₂₀ +y ₀₂

${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 20</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 02</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 11</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

I ₃ =(y ₃₀ +3y) ² +(3y ₂₁ -y ₀₃ ) ²

I ₄ =(y ₃₀ +y ₁₂ ) ² +(y ₂₁ +y ₀₃ ) ²

I ₅ =(y ₃₀ -y ₁₂ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]

+(3y ₂₁ -y ₀₃ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

I ₆ =(y ₂₀ -y ₀₂ )[(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]+

4y ₁₁ (y ₃₀ +y ₁₂ )(y ₂₁ +y ₀₃ )

I ₇ =(3y ₂₁ +y ₀₃ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]+

(y ₃₀ -3y ₁₂ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ].

In a specific implementation, the method also includes calculating the average gray level of the formed components according to the following formula One or more of the standard deviation σ _g of gray levels, the slope κ of red, green and blue, entropy and energy are used as features in the feature set of the classifier:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; g</span>}$

Among them, the probability density function Divide the grayscale into L grayscale levels, h(g) is the number of pixels at grayscale level g in the formed component, M is the total number of pixels in the formed component,

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; =</span>\sqrt{\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; \&kappa;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}^{\mathrm{<span\; class="notranslate">\; 3</span>}}}\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span>$

In a specific implementation, the shaped components include red blood cells, white blood cells, and crystals.

According to an embodiment of the present invention, there is provided a urine formed component classification device, including: a characterization unit configured to convert the average pixel value of the peripheral area of the formed component in the formed component block and the average pixel value of the central area The difference in values is used as a feature in a feature set of a classifier; a classification unit configured to classify urine formed components with said classifier.

In a specific implementation, the characterization unit is further configured to remove the influence of ambient light and defocus noise on the pixel values in the formed component, and calculate the average value of the peripheral area after removing the influence of ambient light and defocus noise The difference between the pixel value and the average pixel value of the central region is used as a feature in the feature set.

In a specific implementation, the characterization unit is further configured to:

For pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. color factor,

${\mathrm{<span\; class="notranslate">\; NT</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}$

max(T _red ), max(T _green ), max(T _blue ) represent the maximum value of the three-color coefficient of red, the maximum value of the green three-color coefficient, and the maximum value of the blue three-color coefficient of the pixel in the formed component, NT _green represents the normalized value of the green three-color coefficient of the pixel;

The difference between the average value of NT _green of each pixel in the outer peripheral area of the formed component and the average value of NT _green of each pixel in the central area is calculated.

In a specific implementation, the characterization unit is further configured to:

For pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

T _gray =0.2989T _red +0.5870T _green +0.1140T _blue

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. Color coefficient, T _gray represents the gray coefficient of the pixel,

${\mathrm{<span\; class="notranslate">\; NT</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; gray</span>}}<span\; class="notranslate">\; )</span>}$

max(T _gray ) represents the maximum value of the gray coefficient of the pixel in the formed component, and NT _gray represents the normalized value of the gray coefficient of the pixel;

The difference between the average value of NT _gray of each pixel in the outer peripheral area of the formed component and the average value of NT _gray of each pixel in the central area is calculated.

In a specific implementation, the characterization unit is further configured to: obtain the gray level co-occurrence matrix of the formed component, and use one or more of the contrast, homogeneity, and energy of the gray level co-occurrence matrix of the formed component features in a feature set used as a classifier, where the contrast, homogeneity, and energy of the gray-level co-occurrence matrix are calculated as follows:

Among them, P(i, j) represents the probability that the values of gray level i and gray level j appear at the same time.

In a specific implementation, the characterization unit is further configured to:

The difference between the average brightness of the pixels in the foreground part of the shaped partition and the average brightness of the pixels in the background part of the shaped partition is used as a feature in the feature set of the classifier.

In a specific implementation, the characterization unit is further configured to use one of the seven Hu distances I ₁ , I ₂ , I ₃ , I ₄ , I ₅ , I ₆ , and I ₇ calculated according to the following process or features from multiple feature sets used as classifiers:

Convert the formed component into a grayscale image, f(x,y) represents the grayscale of the pixel (x,y) in the grayscale image, f(x,y)=0.2989I _red +0.5870I _green +0.1140I _blue , Among them, I _red , I _green , and I _blue represent the pixel values of the pixel (x, y) in the red layer, green layer, and blue layer respectively, and the (p+q) steps are defined as:

m _pq =∫∫x ^p y ^q f(x,y)dxdy p,q=0,1,2,…

(p+q) order center distance is defined as:

μ _pq =∫∫(xx ₀ ) ^p (yy ₀ ) ^q f(x,y)dxdy

in $x_0=\frac{m_{10}}{m_{00}},{\mathrm{the\; y}}_0=\frac{m_{01}}{m_{00}}$

The standard center distance is

in $r+\frac{p+q+2}{2}$

Thus, the seven Hu distances are defined as:

I ₁ =y ₂₀ +y ₀₂

${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 20</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 02</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; 4</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 11</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

I ₃ =(y ₃₀ +3y) ² +(3y ₂₁ -y ₀₃ ) ²

I ₄ =(y ₃₀ +y ₁₂ ) ² +(y ₂₁ +y ₀₃ ) ²

I ₅ =(y ₃₀ -y ₁₂ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]

+(3y ₂₁ -y ₀₃ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

I ₆ =(y ₂₀ -y ₀₂ )[(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]+

4y ₁₁ (y ₃₀ +y ₁₂ )(y ₂₁ +y ₀₃ )

I ₇ =(3y ₂₁ +y ₀₃ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]+

(y ₃₀ -3y ₁₂ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ].

In a specific implementation, the characterization unit is further configured to use the average gray level of the formed components calculated according to the following formula One or more of the standard deviation σ _g of gray levels, the slope κ of red, green and blue, entropy and energy are used as features in the feature set of the classifier:

$\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; g</span>}$

Among them, the probability density function Divide the grayscale into L grayscale levels, h(g) is the number of pixels at grayscale level g in the formed component, M is the total number of pixels in the formed component,

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; =</span>\sqrt{\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; \&kappa;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}^{\mathrm{<span\; class="notranslate">\; 3</span>}}}\underset{\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; )</span>$

In a specific implementation, the shaped components include red blood cells, white blood cells, and crystals.

The inventor observed and found that under the white light of the microscope, some formed components (such as red blood cells) have a thick circle of green light in the periphery except for the dark area in the middle. Therefore, based on the difference between the average pixel value of the peripheral region of the formed component and the average pixel value of the central region of the formed component in the formed subblock, some formed components can be effectively distinguished, especially red blood cells, white blood cells, and crystals. An embodiment of the present invention uses the difference between the average pixel value of the outer peripheral area of the formed sub-block and the average pixel value of the central area in the formed sub-block as the feature set of the classifier, which effectively distinguishes some formed components and improves Accuracy of Urine Formed Component Classification.

The inventor has also discovered some other features that have not been found in the prior art and can be used for the classification of urine formed components, such as features based on gray-scale co-occurrence matrix, features based on the brightness difference of formed components, and features based on Hu distance , features based on histogram metrics, etc. Using one or more of these features as features in the feature set of the classifier can more effectively distinguish the formed components, thereby improving the accuracy of urine formed components classification.

Description of drawings

These and other features and advantages of the present invention will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings.

Fig. 1 shows a flowchart of a method for classifying urine formed components according to an embodiment of the present invention.

FIG. 2 shows a detailed flowchart of using the difference between the average pixel value of the outer peripheral area of the shaped component and the average pixel value of the central area as a feature in the feature set of the classifier according to an embodiment of the present invention.

Fig. 3 shows a block diagram of a urine formed component classification device according to an embodiment of the present invention.

Fig. 4 shows a structural diagram of a urine formed component classification device according to an embodiment of the present invention.

Detailed ways

In the following, various embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Use features based on the green aperture feature as features in the feature set

As shown in Figure 1, the urine formed component classification method 1 according to an embodiment of the present invention includes: in step S1, the average pixel value of the formed component peripheral area and the average pixel value of the central area in the formed block block The difference is used as a feature in the feature set of the classifier; in step S2, the urine formed components are classified by the classifier.

Among them, the peripheral area and the central area are determined by the method of distance transform. Contours of shaped elements can be obtained, for example, in an edge detection process. Indents the outline of the formed element inwards by cells equal to the size of a pixel. When an outline is indented by n spaces, the outline cannot be indented inward. The criterion for indentation that cannot be indented is, for example, the distance between any point on the contour line that is indented by n spaces and the opposite point on the contour line that is indented by n spaces along the direction of indentation If the size is less than or equal to 2 grids, it is considered that it cannot be indented." When the outline cannot be indented, the definition of the part enclosed when the outline of the formed component is indented by n×2/3 grids is the central area, and the part outside the central area of the formed component is defined as the peripheral area.

In another embodiment, the part enclosed when the outline of the formed component is indented by n×1/2 grids is defined as the central area, and the part outside the central area of the formed component is defined as the peripheral area.

As mentioned above, under the white light of the microscope, some formed components (such as red blood cells) have a wide circle of green light around them except for the dark area in the middle. The difference between the average pixel value of the outer peripheral area of the formed component and the average pixel value of the central area reflects this characteristic. Therefore, using the difference between the average pixel value of the outer peripheral area of the formed element and the average pixel value of the central area as a feature in the feature set is helpful for the classification of the formed element.

As shown in Figure 2, in one embodiment, step S1 can be subdivided into the following steps: in step S11, remove the influence of ambient light and defocus noise on the pixel value in the formed component; in step S12, calculate and remove The difference between the average pixel value of the peripheral area and the average pixel value of the central area after the influence of ambient light and defocus noise; in step S13, the difference is used as a feature in the feature set.

Those skilled in the art should understand that step S11 can also be omitted. Because removing the influence of ambient light and out-of-focus noise on the pixel values in the formed components only makes the calculation of the difference more accurate, the average pixel of the peripheral region can still be calculated without removing the influence of ambient light and out-of-focus noise The difference between the value and the average pixel value in the central area, the calculation result can also reflect the evaluation of the above-mentioned green aperture characteristics to a certain extent.

Regarding the calculation of the difference between the average pixel value of the peripheral area and the average pixel value of the central area, two examples are given below.

In one embodiment, for pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$ Formula 1

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. color factor.

${\mathrm{NT}}_{\mathrm{green}}=\frac{T_{\mathrm{green}}}{\max (\max \left(T_{\mathrm{red}}\right),\max \left(T_{\mathrm{green}}\right),\max \left(T_{\mathrm{blue}}\right))}$ Formula 2

Wherein max(T _red ), max(T _green ), max(T _blue ) represent the maximum value of the red trichromatic coefficient, the maximum green trichromatic coefficient, and the blue trichromatic coefficient of the pixel in the formed component. The maximum value, NT _green indicates the normalized value of the green three-color coefficient of the pixel.

Through the above formulas 1 and 2, for each pixel in the formed component, the normalized value NT _green of the green three-color coefficient of the pixel can be calculated. Then, the difference between the average value of NT _green of each pixel in the outer peripheral area of the formed component and the average value of NT _green of each pixel in the central area is calculated. The difference between the average pixel value of the peripheral area and the average pixel value of the central area calculated by using the normalized value of the green three-color coefficient can better reflect the above-mentioned characteristics of the green aperture.

In another embodiment, the inventors found that the difference between the average pixel value of the peripheral area and the average pixel value of the central area calculated by using the gamma normalized value based on the gray scale can also reflect the above-mentioned green halo to a certain extent characteristic.

In this example, for pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$ Formula 3

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

T _gray =0.2989T _red +0.5870T _green +0.1140T _blue

Wherein I _red , I _green , and I _blue respectively represent the pixel values of the pixel in the red layer, green layer, and blue layer, I _gray is the grayscale pixel value of the pixel, and T _red , T _green , T _blue represent respectively The red, green and blue three-color coefficients of the pixel, T _gray represents the gray coefficient of the pixel.

${\mathrm{NT}}_{\mathrm{gray}}=\frac{T_{\mathrm{gray}}}{\max \left(T_{\mathrm{gray}}\right)}$ Formula 4

max(T _gray ) represents the maximum value of the gray coefficient of the pixel in the formed component, and NT _gray represents the normalized value of the gray coefficient of the pixel.

Through the above formulas 3 and 4, for each pixel in the formed component, the gray coefficient normalized value NT _gray of the pixel can be calculated. Then, the difference between the average value of NT _gray of each pixel in the formed peripheral area and the average value of NT _gray of each pixel in the central area is calculated.

Those skilled in the art should understand that, in addition to the difference between the average pixel value of the peripheral area and the average pixel value of the central area calculated by using the normalized value of the green three-color coefficient, the average value of the peripheral area calculated by using the normalized value of the gray coefficient The difference between the pixel value and the average pixel value of the central area, and the difference between the average pixel value of the peripheral area and the average pixel value of the central area calculated by the normalized value of the red and blue coefficients are also helpful for distinguishing the formed components . Therefore, the calculation of the difference between the average pixel value of the peripheral area and the average pixel value of the central area is not limited to the above two embodiments.

Those skilled in the art should understand that although the central area and the peripheral area are defined above as the parts inside and outside the outline after the outline of the formed component is indented by n×1/2 or n×2/3 grids, Other ratios than 2/3 and 1/2 can also be specified. For example, the central area and the peripheral area are defined as the inner and outer parts of the outline after the outline of the formed component is indented by n×3/4 grids.

Those skilled in the art should understand that the central area and the peripheral area can be defined according to the inner and outer parts of the contour after the contour of the formed component is indented by n×2/3 grids, and the average pixels of the peripheral area can be calculated accordingly value and the first difference of the average pixel value of the central region. Then define the central area and the outer peripheral area according to the inner and outer parts of the outline after the outline of the formed component is indented inward by n×1/2 grids, and calculate the average pixel value of the outer peripheral area and the average pixel value of the central area accordingly The second worst value. Both the first difference and the second difference are taken as two features in the feature set. In addition, when the difference between the average pixel value of the peripheral area and the average pixel value of the central area is calculated by using both the normalized value of the green tricolor coefficient and the normalized value of the gray coefficient, four pixels are added to the feature set feature.

Use gray level co-occurrence matrix based features as features in feature set

The inventors of the present invention found that some formed components, especially red blood cells, white blood cells, and crystals, can also be effectively distinguished by using some features based on the gray scale co-occurrence matrix.

According to an embodiment of the present invention, the gray level co-occurrence matrix of the formed components is obtained. For a specific formed component, its gray level co-occurrence matrix can be determined by existing technology. Then, one or more of the contrast, homogeneity, energy of the gray level co-occurrence matrix of the formed components is used as features in the feature set of the classifier. The contrast, homogeneity, and energy of the gray-level co-occurrence matrix are calculated as follows:

Formula 5

Formula 6

Formula 7

Among them, P(i, j) represents the probability that the values of gray level i and gray level j appear at the same time.

One or more of the contrast, homogeneity, and energy of the gray-level co-occurrence matrix of the formed components is used as the feature set of the classifier, which can improve the effectiveness of distinguishing the formed components and improve the classification accuracy.

Use features based on the contrast of shaped partitions as features in the feature set

The inventors found that for some formed components, the brightness of the foreground part (the formed component itself) and the background part of the block under the microscope are very different, which can also help to distinguish some formed components. Therefore, according to one embodiment of the present invention, the difference between the average brightness of the pixels in the foreground part of the formed partition and the average brightness of the pixels in the background part of the formed partition is used as a feature in the feature set of the classifier. This can also improve the effectiveness of distinguishing formed elements (especially red blood cells, white blood cells, crystals) and improve classification accuracy.

Use Hu distance based features as features in feature set

The inventors found that, for some formed components, its distance is also very different from other formed components. Therefore, one or more of the seven Hu distances I ₁ , I ₂ , I ₃ , I ₄ , I ₅ , I ₆ , and I ₇ calculated according to the following process are used as features in the feature set of the classifier , It can also improve the effectiveness of distinguishing formed components (especially red blood cells, white blood cells, and crystals), and improve the classification accuracy.

First, convert the formed component into a grayscale image, f(x,y) represents the grayscale of the pixel (x,y) in the grayscale image, f(x,y)=0.2989I _red +0.5870I _green +0.1140I _blue , where I _red , I _green , and I _blue represent the pixel values of the pixel (x, y) in the red layer, green layer, and blue layer respectively, and the (p+q) step is defined as:

m _pq =∫∫x ^p y ^q f(x,y)dxdy p,q=0,1,2,… Formula 8

(p+q) order center distance is defined as:

μ _pq =∫∫(xx ₀ ) ^p (yy ₀ ) ^q f(x,y)dxdy Formula 9

in $x_0=\frac{m_{10}}{m_{00}},{\mathrm{the\; y}}_0=\frac{m_{01}}{m_{00}}.$

The standard center distance is Formula 10

in $r+\frac{p+q+2}{2}$

Thus, the seven Hu distances are defined as:

I ₁ =y ₂₀ +y ₀₂ Formula 11

$I_2={({\mathrm{the\; y}}_{20}+{\mathrm{the\; y}}_{02})}^2+{4\mathrm{the\; y}}_{11}^2$ Formula 12

I ₃ =(y ₃₀ +3y) ² +(3y ₂₁ -y ₀₃ ) ² Formula 13

I ₄ =(y ₃₀ +y ₁₂ ) ² +(y ₂₁ +y ₀₃ ) ² Formula 14

I ₅ =(y ₃₀ -y ₁₂ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]

+(3y ₂₁ -y ₀₃ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

Formula 15

I ₆ =(y ₂₀ -y ₀₂ )[(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]+ Formula 16

4y ₁₁ (y ₃₀ +y ₁₂ )(y ₂₁ +y ₀₃ )

I ₇ =(3y ₂₁ +y ₀₃ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² −3(y ₂₁ +y ₀₃ ) ² ]+ Formula 17.

(y ₃₀ -3y ₁₂ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

Use features based on histogram metrics as features in a feature set

The inventors found that for some formed components, some histogram measures are also very different from other formed components. Therefore, the histogram measure of the formed components (including the average gray level) calculated according to the following formula One or more of the standard deviation σ _g of the gray level, the slope κ of red, green and blue, entropy and energy) are used as features in the feature set of the classifier, which can also improve the distinction between formed components (especially red blood cells, white blood cells, and crystals) to improve classification accuracy.

$\stackrel{\&OverBar;}{g}=\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}P\left(g\right)\&CenterDot;g$ Formula 18

Among them, the probability density function The grayscale is divided into L grayscale levels. For example, grayscales within a certain grayscale range are divided into one grayscale level. L can be determined based on empirical values or through limited experiments and simulations as needed, and usually L cannot be too small. h(g) is the number of pixels in the formed component at gray level g, and M is the total number of pixels in the formed component.

${\mathrm{\&sigma;}}_g=\sqrt{\underset{g-0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^2\&Center\; Dot;P\left(g\right)}$ Formula 19

$\mathrm{\&kappa;}=\frac{1}{{\mathrm{\&sigma;}}_g^3}\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^3\&CenterDot;P\left(g\right)$ Formula 20

Formula 21

Formula 22.

A device according to one embodiment of the invention

As shown in FIG. 3 , the urine formed component classification device 2 according to an embodiment of the present invention includes a characterization unit 201 and a classification unit 202 . The characterizing unit 201 is configured to use the difference between the average pixel value of the outer peripheral area of the formed segment and the average pixel value of the central area in the formed segment block as a feature in the feature set of the classifier. The classification unit 202 is configured to classify the urine formed components with the classifier. The apparatus shown in FIG. 3 can be realized by software, hardware (such as an integrated circuit, FPGA, etc.), or a combination of software and hardware.

In one embodiment, the characterization unit 201 may be further configured to remove the influence of ambient light and defocus noise on the pixel values in the formed components, and calculate the The difference between the average pixel value and the average pixel value of the central region is used as a feature in the feature set.

In one embodiment, the characterization unit 201 may be further configured as:

For pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$ Formula 1

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

Wherein I _red , I _green , and I _blue represent the pixel values of the pixel in the red layer, green layer, and blue layer respectively, and T _red , T _green , and T _blue represent the red, green, and blue three-dimensional values of the pixel respectively. color factor,

${\mathrm{NT}}_{\mathrm{green}}=\frac{T_{\mathrm{green}}}{\max (\max \left(T_{\mathrm{red}}\right),\max \left(T_{\mathrm{green}}\right),\max \left(T_{\mathrm{blue}}\right))}$ Formula 2

Among them, max(T _red ), max(T _green ), max(T _blue ) represent the maximum value of red trichromatic coefficients, the maximum green trichromatic coefficients, and the blue trichromatic coefficients of pixels in the formed components. The maximum value of the coefficient, NT _green represents the normalized value of the green three-color coefficient of the pixel;

The difference between the average value of NT _green of each pixel in the outer peripheral area of the formed component and the average value of NT _green of each pixel in the central area is calculated.

In one embodiment, the characterization unit 201 may be further configured as:

For pixels in the formed component, compute

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$ Formula 3

${\mathrm{<span\; class="notranslate">\; T</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; red</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; green</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; blue</span>}}}$

T _gray =0.2989T _red +0.5870T _green +0.1140T _blue

Wherein I _red , I _green , and I _blue respectively represent the pixel values of the pixel in the red layer, green layer, and blue layer, I _gray is the grayscale pixel value of the pixel, and T _red , T _green , T _blue represent respectively The red, green and blue three-color coefficients of the pixel, T _gray represents the gray coefficient of the pixel,

${\mathrm{NT}}_{\mathrm{gray}}=\frac{T_{\mathrm{gray}}}{\max \left(T_{\mathrm{gray}}\right)}$ Formula 4

max(T _gray ) represents the maximum value of the gray coefficient of the pixel in the formed component, and NT _gray represents the normalized value of the gray coefficient of the pixel;

The difference between the average value of NT _gray of each pixel in the outer peripheral area of the formed component and the average value of NT _gray of each pixel in the central area is calculated.

In one embodiment, the characterization unit 201 is further configured to:

Obtain the gray level co-occurrence matrix of the formed component, and use one or more of the contrast, homogeneity, and energy of the gray level co-occurrence matrix of the formed component as the feature set of the classifier, wherein the gray level co-occurrence matrix Contrast, homogeneity, energy are calculated as follows:

Formula 5

Formula 6

Formula 7

Among them, P(i, j) represents the probability that the values of gray level i and gray level j appear at the same time.

In one embodiment, the characterization unit 201 can also be configured to use the difference between the average brightness of pixels in the foreground part of the formed block and the average brightness of the pixels in the background of the formed block as the feature set of the classifier Characteristics.

In one embodiment, the characterization unit 201 can also be configured to use the seven Hu distances I ₁ , I ₂ , I ₃ , I ₄ , I ₅ , I ₆ , and I ₇ calculated according to the following process One or more features in the feature set used as features in the classifier:

Convert the formed component into a grayscale image, f(x, _y ) represents the grayscale of the pixel (x, _y ) in the grayscale image, f(x,y)=0.2989I _red +0.5870I _green +0.1140I _blue , Among them, I _red , I _green , and I _blue represent the pixel values of the pixel (x, y) in the red layer, green layer, and blue layer respectively, and the (p+q) steps are defined as:

m _pq =∫∫x ^p y ^q f(x,y)dxdy p,q=0,1,2,… Formula 8

(p+q) order center distance is defined as:

μ _pq =∫∫(xx ₀ ) ^p (yy ₀ ) ^q f(x,y)dxdy Formula 9

in $x_0=\frac{m_{10}}{m_{00}},{\mathrm{the\; y}}_0=\frac{m_{01}}{m_{00}}$

The standard center distance is Formula 10

in $r+\frac{p+q+2}{2}$

Thus, the seven Hu distances are defined as:

I ₁ =y ₂₀ +y ₀₂ Formula 11

$I_2={({\mathrm{the\; y}}_{20}+{\mathrm{the\; y}}_{02})}^2+{4\mathrm{the\; y}}_{11}^2$ Formula 12

I ₃ =(y ₃₀ +3y) ² +(3y ₂₁ -y ₀₃ ) ² Formula 13

I ₄ =(y ₃₀ +y ₁₂ ) ² +(y ₂₁ +y ₀₃ ) ² Formula 14

I ₅ =(y ₃₀ -y ₁₂ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ] Formula 15

+(3y ₂₁ -y ₀₃ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

I ₆ =(y ₂₀ -y ₀₂ )[(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]+ Formula 16

4y ₁₁ (y ₃₀ +y ₁₂ )(y ₂₁ +y ₀₃ )

I ₇ =(3y ₂₁ +y ₀₃ )(y ₃₀ +y ₁₂ )[(y ₃₀ +y ₁₂ ) ² -3(y ₂₁ +y ₀₃ ) ² ]+ Formula 17

(y ₃₀ -3y ₁₂ )(y ₂₁ +y ₃₀ )[3(y ₃₀ +y ₁₂ ) ² -(y ₂₁ +y ₀₃ ) ² ]

In one embodiment, the characterization unit 201 can also be configured to use the average gray level of the formed components calculated according to the following formula One or more of the standard deviation σ _g of gray levels, the slope κ of red, green and blue, entropy, and energy are used as features in the feature set of the classifier

$\stackrel{\&OverBar;}{g}=\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}P\left(g\right)\&Center\; Dot;g$ Formula 18

Among them, the probability density function The grayscale is divided into L grayscale levels. For example, grayscales within a certain grayscale range are divided into one grayscale level. L can be determined according to needs, but L cannot be too small. h(g) is the number of pixels in the formed component at gray level g, and M is the total number of pixels in the formed component.

${\mathrm{\&sigma;}}_g=\sqrt{\underset{g-0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^2\&CenterDot;P\left(g\right)}$ Formula 19

$\mathrm{\&kappa;}=\frac{1}{{\mathrm{\&sigma;}}_g^3}\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^3\&Center\; Dot;P\left(g\right)$ Formula 20

Formula 21

Formula 22

In one embodiment, said formed elements include red blood cells, white blood cells, crystals.

Fig. 4 shows a urine formed component classification device 3 according to one embodiment of the present invention. The device may include a memory 301 and a processor 302 . The memory 301 is used to store executable instructions. The processor 302 is configured to perform operations performed by each unit in the device 2 according to the executable instructions stored in the memory.

In addition, an embodiment of the present invention also provides a machine-readable medium on which executable instructions are stored, and when the executable instructions are executed, the machine executes the operations performed by the processor 302 .

Those skilled in the art should understand that various variations and modifications can be made to the above embodiments without departing from the essence of the invention, therefore, the protection scope of the present invention should be defined by the appended claims.

Claims (20)

1. a urinary formed element sorting technique (1), comprising:

By the difference of the average pixel value of visible component outer region and the average pixel value of central area in visible component block, as the feature (S1) in the feature set of sorter;

With described sorter to urinary formed element classification (S2).

2. urinary formed element sorting technique (1) according to claim 1, the step (S1) wherein the difference of the average pixel value of visible component outer region and the average pixel value of central area being used as the feature in the feature set of sorter comprising:

Removal surround lighting and noise out of focus are on the impact (S11) of the pixel value in visible component;

Calculate the difference (S12) of the average pixel value of outer region after the impact of removing surround lighting and noise out of focus and the average pixel value of central area;

Described difference is used as the feature (S13) in described feature set.

3. urinary formed element sorting technique (1) according to claim 2, wherein

Removal surround lighting and the step (S11) of noise out of focus on the impact of the pixel value in visible component comprising: for the pixel in visible component, calculate

$T_{\mathrm{red}}=\frac{I_{\mathrm{red}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{blue}}=\frac{I_{\mathrm{blue}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

Wherein I _red, I _green, I _bluerepresent the pixel value of described pixel at red color layer, green layer, cyan coloring layer respectively, T _red, T _green, T _bluerepresent the redness of described pixel, green, blue trichromatic coefficient respectively,

${\mathrm{NT}}_{\mathrm{green}}=\frac{T_{\mathrm{green}}}{\max (\max \left(T_{\mathrm{red}}\right),\max \left(T_{\mathrm{green}}\right),\max \left(T_{\mathrm{blue}}\right))}$

Max (T _red), max (T _green), max (T _blue) represent the trichromatic coefficient maximal value of the redness of pixel in visible component, green trichromatic coefficient maximal value, blue trichromatic coefficient maximal value, NT _greenrepresent the green trichromatic coefficient normalized value of described pixel;

The step (S12) calculating the difference of the average pixel value of outer region after the impact of removing surround lighting and noise out of focus and the average pixel value of central area comprising: the NT calculating each pixel of visible component outer region _greenmean value and the NT of each pixel of central area _greenthe difference of mean value.

4. urinary formed element sorting technique (1) according to claim 2, wherein

Removal surround lighting and the step (S11) of noise out of focus on the impact of the pixel value in visible component comprising: for the pixel in visible component, calculate

$T_{\mathrm{red}}=\frac{I_{\mathrm{red}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{blue}}=\frac{I_{\mathrm{blue}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

T _gray=0.2989T _red+0.5870T _green+0.1140T _blue

Wherein I _red, I _green, I _bluerepresent the pixel value of described pixel at red color layer, green layer, cyan coloring layer respectively, T _red, T _green, T _bluerepresent the redness of described pixel, green, blue trichromatic coefficient respectively, T _grayrepresent the grey coefficients of described pixel,

${\mathrm{NT}}_{\mathrm{gray}}=\frac{T_{\mathrm{gray}}}{\max \left(T_{\mathrm{gray}}\right)}$

Max (T _gray) represent the grey coefficients maximal value of pixel in visible component, NT _grayrepresent the grey coefficients normalized value of this pixel;

The step (S12) calculating the difference of the average pixel value of outer region after the impact of removing surround lighting and noise out of focus and the average pixel value of central area comprising: the NT calculating each pixel of visible component outer region _graymean value and the NT of each pixel of central area _graythe difference of mean value.

5. urinary formed element sorting technique (1) according to claim 1, described method also comprises:

Obtain the gray level co-occurrence matrixes of visible component, and one or more as the feature in the feature set of sorter using in the contrast of the gray level co-occurrence matrixes of visible component, homogeney, energy, wherein the contrast of gray level co-occurrence matrixes, homogeney, energy calculate as follows:

Wherein P(i, j) represent gray scale to be i and gray scale be the probability that the value of j occurs simultaneously.

6. urinary formed element sorting technique (1) according to claim 1, described method also comprises:

The difference of the average luminance of pixels of background parts in the average luminance of pixels of prospect part in visible component block and visible component block is used as the feature in the feature set of sorter.

7. urinary formed element sorting technique (1) according to claim 1, described method also comprises 7 the Hu distance I will gone out according to following process computation ₁, I ₂, I ₃, I ₄, I ₅, I ₆, I ₇in one or more be used as sorter feature sets in feature:

Convert visible component to gray level image, f (x, y) represents the gray scale of pixel (x, y) in gray level image, f (x, y)=0.2989I _red+ 0.5870I _green+ 0.1140I _blue, wherein I _red, I _green, I _bluerepresent the pixel value of described pixel (x, y) at red color layer, green layer, cyan coloring layer respectively, its (p+q) rank are apart from being defined as:

m _pq=∫∫x ^py ^qf(x,y)dxdy p,q=0,1,2,…

(p+q) rank centre distance is defined as:

μ _pq=∫∫(x-x ₀) ^p(y-y ₀) ^qf(x,y)dxdy

Wherein $x_0=\frac{m_{10}}{m_{00}},y_0=\frac{m_{01}}{m_{00}}$

Reference center distance is

Wherein $r+\frac{p+q+2}{2}$

Thus, 7 Hus are apart from being defined as:

I ₁=y ₂₀+y ₀₂

$I_2={(y_{20}+y_{02})}^2+{4y}_{11}^2$

I ₃=(y ₃₀+3y) ²+(3y ₂₁-y ₀₃) ²

I ₄=(y ₃₀+y ₁₂) ²+(y ₂₁+y ₀₃) ²

I ₅=(y ₃₀-y ₁₂)(y ₃₀+y ₁₂)[(y ₃₀+y ₁₂) ²-3(y ₂₁+y ₀₃) ²]

+(3y ₂₁-y ₀₃)(y ₂₁+y ₃₀)[3(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]

I ₆=(y ₂₀-y ₀₂)[(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]+

4y ₁₁(y ₃₀+y ₁₂)(y ₂₁+y ₀₃)

I ₇=(3y ₂₁+y ₀₃)(y ₃₀+y ₁₂)[(y ₃₀+y ₁₂) ²-3(y ₂₁+y ₀₃) ²]+

(y ₃₀-3y ₁₂)(y ₂₁+y ₃₀)[3(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]。

8. urinary formed element sorting technique (1) according to claim 1, described method also comprises the average gray level of the visible component calculated according to the following formula the standard deviation sigma of gray level _g, the gradient κ of RGB look, feature in one or more feature sets being used as sorter in entropy and energy:

$\stackrel{\&OverBar;}{g}=\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}P\left(g\right)\&CenterDot;g$

Wherein, probability density function gray scale be divided in L grey level, h (g) is the pixel count being in grey level g in visible component, and M is the sum of all pixels in visible component,

${\mathrm{\&sigma;}}_g=\sqrt{\underset{g-0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^2\&CenterDot;P\left(g\right)}$

$\mathrm{\&kappa;}=\frac{1}{{\mathrm{\&sigma;}}_g^3}\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^3\&CenterDot;P\left(g\right)$

9. urinary formed element sorting technique (1) according to claim 1, wherein said visible component comprises red blood cell, leucocyte, crystallization.

10. a urinary formed element sorter (2), comprising:

Characterization unit (201), is configured to the difference of the average pixel value of visible component outer region and the average pixel value of central area in visible component block, as the feature in the feature set of sorter;

Taxon (202), is configured to classify to urinary formed element with described sorter.

11. urinary formed element sorters (2) according to claim 10, wherein said characterization unit (201) is configured to remove surround lighting and noise out of focus to the impact of the pixel value in visible component further, calculate the difference of the average pixel value of outer region after the impact of removing surround lighting and noise out of focus and the average pixel value of central area, described difference is used as the feature of described feature set.

12. urinary formed element sorters (2) according to claim 11, wherein said characterization unit (201) is configured to further:

For the pixel in visible component, calculate

$T_{\mathrm{red}}=\frac{I_{\mathrm{red}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{blue}}=\frac{I_{\mathrm{blue}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

Wherein I _red, I _green, I _bluerepresent the pixel value of described pixel at red color layer, green layer, cyan coloring layer respectively, T _red, T _green, T _bluerepresent the redness of described pixel, green, blue trichromatic coefficient respectively,

${\mathrm{NT}}_{\mathrm{green}}=\frac{T_{\mathrm{green}}}{\max (\max \left(T_{\mathrm{red}}\right),\max \left(T_{\mathrm{green}}\right),\max \left(T_{\mathrm{blue}}\right))}$

Max (T _red), max (T _green), max (T _blue) represent the trichromatic coefficient maximal value of the redness of pixel in visible component, green trichromatic coefficient maximal value, blue trichromatic coefficient maximal value, NT _greenrepresent the green trichromatic coefficient normalized value of described pixel;

Calculate the NT of each pixel of visible component outer region _greenmean value and the NT of each pixel of central area _greenthe difference of mean value.

13. urinary formed element sorters (2) according to claim 11, wherein said characterization unit (201) is configured to further:

For the pixel in visible component, calculate

$T_{\mathrm{red}}=\frac{I_{\mathrm{red}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{green}}=\frac{I_{\mathrm{green}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

$T_{\mathrm{blue}}=\frac{I_{\mathrm{blue}}}{I_{\mathrm{red}}+I_{\mathrm{green}}+I_{\mathrm{blue}}}$

T _gray=0.2989T _red+0.5870T _green+0.1140T _blue

Wherein I _red, I _green, I _bluerepresent the pixel value of described pixel at red color layer, green layer, cyan coloring layer respectively, T _red, T _green, T _bluerepresent the redness of described pixel, green, blue trichromatic coefficient respectively, T _grayrepresent the grey coefficients of described pixel,

${\mathrm{NT}}_{\mathrm{gray}}=\frac{T_{\mathrm{gray}}}{\max \left(T_{\mathrm{gray}}\right)}$

Max (T _gray) represent the grey coefficients maximal value of pixel in visible component, NT _grayrepresent the grey coefficients normalized value of this pixel;

Calculate the NT of each pixel of visible component outer region _graymean value and the NT of each pixel of central area _graythe difference of mean value.

14. urinary formed element sorters (2) according to claim 10, wherein said characterization unit (201) is also configured to:

Obtain the gray level co-occurrence matrixes of visible component, and by the feature in the one or more feature sets being used as described sorter in the contrast of the gray level co-occurrence matrixes of visible component, homogeney, energy, wherein the contrast of gray level co-occurrence matrixes, homogeney, energy calculate as follows:

Wherein P(i, j) represent gray scale to be i and gray scale be the probability that the value of j occurs simultaneously.

15. urinary formed element sorters (2) according to claim 10, wherein said characterization unit (201) is also configured to:

The difference of the average luminance of pixels of background parts in the average luminance of pixels of prospect part in visible component block and visible component block is used as the feature in the feature set of described sorter.

16. urinary formed element sorters (2) according to claim 10, wherein said characterization unit (201) is also configured to 7 the Hu distance I will gone out according to following process computation ₁, I ₂, I ₃, I ₄, I ₅, I ₆, I ₇in one or more be used as described sorter feature sets in feature:

Convert visible component to gray level image, f (x, y) represents the gray scale of pixel (x, y) in gray level image, f (x, y)=0.2989I _red+ 0.5870I _green+ 0.1140I _blue, wherein I _red, I _green, I _bluerepresent the pixel value of described pixel (x, y) at red color layer, green layer, cyan coloring layer respectively, its (p+q) rank are apart from being defined as:

m _pq=∫∫x ^py ^qf(x,y)dxdy p,q=0,1,2,…

(p+q) rank centre distance is defined as:

μ _pq=∫∫(x-x ₀) ^p(y-y ₀) ^qf(x,y)dxdy

Wherein $x_0=\frac{m_{10}}{m_{00}},y_0=\frac{m_{01}}{m_{00}}$

Reference center distance is

Wherein $r+\frac{p+q+2}{2}$

Thus, 7 Hus are apart from being defined as:

I ₁=y ₂₀+y ₀₂

$I_2={(y_{20}+y_{02})}^2+{4y}_{11}^2$

I ₃=(y ₃₀+3y) ²+(3y ₂₁-y ₀₃) ²

I ₄=(y ₃₀+y ₁₂) ²+(y ₂₁+y ₀₃) ²

I ₅=(y ₃₀-y ₁₂)(y ₃₀+y ₁₂)[(y ₃₀+y ₁₂) ²-3(y ₂₁+y ₀₃) ²]

+(3y ₂₁-y ₀₃)(y ₂₁+y ₃₀)[3(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]

I ₆=(y ₂₀-y ₀₂)[(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]+

4y ₁₁(y ₃₀+y ₁₂)(y ₂₁+y ₀₃)

I ₇=(3y ₂₁+y ₀₃)(y ₃₀+y ₁₂)[(y ₃₀+y ₁₂) ²-3(y ₂₁+y ₀₃) ²]+

(y ₃₀-3y ₁₂)(y ₂₁+y ₃₀)[3(y ₃₀+y ₁₂) ²-(y ₂₁+y ₀₃) ²]。

17. urinary formed element sorters (2) according to claim 10, wherein said characterization unit (201) is also configured to the average gray level of the visible component calculated according to the following formula the standard deviation sigma of gray level _g, the gradient κ of RGB look, feature in one or more feature sets being used as described sorter in entropy and energy:

$\stackrel{\&OverBar;}{g}=\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}P\left(g\right)\&CenterDot;g$

Wherein, probability density function gray scale be divided in L grey level, h (g) is the pixel count being in grey level g in visible component, and M is the sum of all pixels in visible component,

${\mathrm{\&sigma;}}_g=\sqrt{\underset{g-0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^2\&CenterDot;P\left(g\right)}$

$\mathrm{\&kappa;}=\frac{1}{{\mathrm{\&sigma;}}_g^3}\underset{g=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{(g-\stackrel{\&OverBar;}{g})}^3\&CenterDot;P\left(g\right)$

18. urinary formed element sorters (2) according to claim 10, wherein said visible component comprises red blood cell, leucocyte, crystallization.

19. 1 kinds of urinary formed element sorting devices (3), comprising:

Storer (301), for stores executable instructions;

Processor (302), for the executable instruction stored according to described storer, enforcement of rights requires the operation performed by any one claim in 1-9.

20. 1 kinds of machine readable medias, it stores executable instruction, when described executable instruction is performed, makes the operation performed by any one claim in machine enforcement of rights requirement 1-9.