CN101482927A - Foreign fiber fuzzy classification system and method based on automatic vision detection - Google Patents

Abstract

The invention relates to a cotton foreign fiber classification system and method thereof. Multi-featured vector is formed by selecting color feature, shape feature and texture feature, and corresponding vague classifier will be structured, therefore the class recognition of foreign fiber is achieved. The invention detects the content of foreign fiber in the cotton with convenience, rapidity and precision, and it also classifies the cotton. The invention make the cotton seller reduce the content of foreign fiber in cotton.

Description

Foreign fiber fuzzy classification system and method thereof based on the automatic vision detection

Technical field

The present invention relates to cotton foreign fiber classification field, be specifically related to a kind of cotton foreign fiber fuzzy classification system and method thereof that detects based on automatic vision.

Background technology

Foreign fiber in the cotton is meant at cotton picking, teds, purchases, stores, in the transportation, process, sneaking into has non-cotton fiber and the color fibre that has a strong impact on to cotton and Products Quality thereof in the cotton, be commonly called as " three ", mainly comprise and compile Polypropylence Sheet, cloth, hair, the rope made of hemp, polypropylene fibre silk and chicken feather etc.

Though the content of foreign fiber in gined cotton is few, and is serious to the quality influence of wollen fabrics, will influence yarn strength in case sneak into and participate in weaving, and influence is painted when dyeing, can cause heavy economic losses to cotton textiles industry.General solution is the content that reduces foreign fiber by picking at present.Think but the expert of a lot of cotton industries is consistent, reduce the content of foreign fiber by the mode of picking (no matter be manually or machinery), always passive and passive; More active and positive way should be to pick up from the source, cut off the approach that foreign fiber is sneaked into, and allow cotton grower, cotton purchase person and processing enterprise all consciously take precautions against foreign fiber and sneak into.And just become the key point of problem at the content of profiled fibre that unginned cotton is purchased or how bargain link is accurately analyzed in the unginned cotton.

Domestic detection to cotton foreign fiber content at present only limits to gined cotton.General way is: the gined cotton bale of cotton that cotton ginning mill produces is taken a sample, deliver to examination of fibers mechanism then and keep sample, and analyze the content of determining foreign fiber.This method is not also used in unginned cotton purchase link, therefore at present also can't utilize content of profiled fibre to influence the purchasing price of unginned cotton.

AVI (Automated Visual Inspection, automatic vision detects) be to utilize optical device to carry out the noncontact perception, automatically obtain the various visual characteristics of target to be measured, and it is analyzed and judges, so that reach the purpose that some designated parameter is detected.The AVI technology has been carried out successfully utilization in fields such as industrial part defects detection, pcb board defects detection, food safety detection, fruit gradings, and has reached good effect.And utilize the AVI technology to carry out the on-line metering of cotton foreign fiber content, yet there are no report in the document at home and abroad.

Utilize the AVI technology to carry out the on-line metering of cotton foreign fiber content, key is that the foreign fiber target that splits is classified, and the accuracy of classification will directly influence the follow-up gage work that carries out.Feature extraction and tagsort are the cores of AVI technology, and main task is to produce the proper vector row mode identification of going forward side by side.Mainly contain color characteristic extraction, Shape Feature Extraction, texture feature extraction etc. about feature extracting methods at present; The sorter that is used for pattern-recognition comprises sorter based on probability statistics, based on the sorter of artificial neural network, based on fuzzy classification device of fuzzy mathematics etc.

Although the research for feature extraction and pattern-recognition has obtained a lot of achievements at present, also there are a lot of achievements in actual production, to carry out successful application, also there is not a kind of method in common can solve all pattern recognition problem at present.More because the foreign fiber kind in the cotton is numerous and diverse, color is different, be difficult to find a kind of method of suitable maturation that it is accurately classified.

Summary of the invention

The purpose of this invention is to provide a kind of cotton foreign fiber fuzzy classification method and system based on AVI, this method selects color characteristic, shape facility and textural characteristics to constitute many proper vectors, and make up corresponding fuzzy classification device respectively, carry out then integrated, thereby realize the kind identification of foreign fiber is classified to described foreign fiber.

In order to reach above purpose, the invention provides a kind of cotton foreign fiber fuzzy classification method based on AVI, said method comprising the steps of:

S1 carries out feature selecting to color characteristic, shape facility and the textural characteristics of foreign fiber target to be identified, obtains the three and has the proper vector that classification is worth separately;

S2, color feature vector, the shape facility vector sum texture feature vector of the foreign fiber target to be identified of extraction S1 gained;

S3, utilize resulting color feature vector among the S2, shape facility vector sum texture feature vector, obtain degree of membership color level component, degree of membership shape level component and degree of membership texture stages component that foreign fiber target to be identified belongs to different foreign fiber classifications;

S4 is weighted summation with described degree of membership color level component, degree of membership shape level component and degree of membership texture stages component, obtains the comprehensive degree of membership that target to be identified belongs to different foreign fiber classifications;

S5 belongs to the comprehensive degree of membership maximizing of different foreign fiber classifications to the foreign fiber target to be identified that obtains among the S4, and the classification that then has maximum comprehensive degree of membership is exactly the maximum possible classification of foreign fiber target to be identified.

Wherein, among the described step S1, selected color feature vector with maximum classification value comprises the average of redness, green, blueness and brightness, the standard deviation of three components of RGB; Selected shape facility vector with maximum classification value comprises form factor, expansion ratio and Euler's numbers; Selected texture feature vector with maximum classification value comprises standard deviation and contrast.

Wherein, utilize resulting color characteristic among the S2, the preliminary classification model of representing by following formula obtains the degree of membership color level component that foreign fiber target to be identified belongs to different foreign fiber classifications:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Color}}=f(\stackrel{\‾}{R},\stackrel{\‾}{G},\stackrel{\‾}{B},\stackrel{\‾}{I},{\mathrm{\σ}}_{\mathrm{RGB}})$

Wherein, R, G, B, I, σ _RGBRepresent redness, green, blueness and the brightness average of target to be identified and the standard deviation of three components of RGB respectively;

Be foreign fiber target to be identified and belong to the degree of membership color level component of different foreign fiber classifications.

Wherein, utilize resulting shape facility among the S2, the preliminary classification model of representing by following formula obtains the degree of membership shape level component that foreign fiber target to be identified belongs to different foreign fiber classifications:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}=g(S,\mathrm{Ex},\mathrm{Eu})$

Wherein, S, Ex, Eu represent form factor, expansion ratio and the Euler's numbers of target to be identified respectively;

Be foreign fiber target to be identified and belong to the degree of membership shape level component of different foreign fiber classifications.

Wherein, utilize resulting textural characteristics among the S2, the preliminary classification model of representing by following formula obtains the degree of membership texture stages component that foreign fiber target to be identified belongs to different foreign fiber classifications:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}=h({\mathrm{\σ}}_{\mathrm{bb}},W_c)$

Wherein, σ _Bb, W _cStandard deviation and the contrast of representing target to be identified respectively;

Be foreign fiber target to be identified and belong to the degree of membership texture stages component of different foreign fiber classifications.

Wherein, among the described step S4, utilize following formula to calculate the comprehensive degree of membership that foreign fiber target to be identified belongs to different foreign fiber classifications:

$\stackrel{\→}{\mathrm{\μ}}={\stackrel{\→}{w}}^{\mathrm{Color}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Color}}+{\stackrel{\→}{w}}^{\mathrm{Shape}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}+{\stackrel{\→}{w}}^{\mathrm{Texture}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}$

Wherein,

Be the weight of degree of membership color level component,

Be the weight of degree of membership shape level component,

Weight for degree of membership texture stages component; Be degree of membership color level component,

Be degree of membership shape level component,

Be degree of membership texture stages component;

Be the comprehensive degree of membership that foreign fiber target to be identified belongs to a different category.

Wherein, in step S5, the method that foreign fiber target to be identified is belonged to the comprehensive degree of membership maximizing of different foreign fiber classifications is:

$\mathrm{FFClass}=\arg \max \left(\stackrel{\→}{\mathrm{\μ}}\right|\stackrel{\→}{\mathrm{\μ}}=\left[{\mathrm{\μ}}_A{\mathrm{\μ}}_B{\mathrm{\μ}}_C{\mathrm{\μ}}_D{\mathrm{\μ}}_E{\mathrm{\μ}}_F\right])$

Wherein,

The comprehensive degree of membership vector that belongs to a different category for foreign fiber target to be identified, arg represents to get the class code of max function gained maximal value representative but not maximal value itself, and FFClass is the class code of the foreign fiber that has maximum comprehensive degree of membership in the vector.

The present invention also provides a kind of cotton foreign fiber fuzzy classification system that detects based on automatic vision, and described system comprises:

Characteristic extracting module is used to carry out the extraction of color characteristic, shape facility and textural characteristics;

Elementary fuzzy classification module links to each other with characteristic extracting module, is used to obtain degree of membership color level component, shape level component and the texture stages component that foreign fiber target to be identified belongs to a different category;

Comprehensive degree of membership computing module links to each other with elementary fuzzy classification module, is used to calculate the comprehensive degree of membership that foreign fiber target to be identified belongs to a different category;

The foreign fiber fuzzy sort module, link to each other with comprehensive degree of membership computing module, be used for that resulting foreign fiber target to be identified is belonged to different classes of comprehensive degree of membership and ask for maximal value, the classification with maximum comprehensive degree of membership is exactly the maximum possible classification of target to be identified.

Wherein, described color characteristic comprises the average of redness, green, blueness, brightness and the standard deviation of three components of RGB; Shape facility comprises form factor, expansion ratio and Euler's numbers; Textural characteristics comprises standard deviation and contrast.

Wherein, described elementary fuzzy classification module comprise elementary fuzzy classification device based on color characteristic, based on the elementary fuzzy classification device of shape facility with based on the elementary fuzzy classification device of textural characteristics.

Wherein, described foreign fiber classification comprises plastics dry goods, cloth class, rope made of hemp class, hair class, polypropylene fibre silk class and chicken feather class.By system extension, can discern how new foreign fiber classification.

Wherein, described comprehensive degree of membership computing module comprises six comprehensive degree of membership counters, is respectively applied for the comprehensive degree of membership of calculating plastics dry goods, cloth class, rope made of hemp class, hair class, polypropylene fibre silk class and chicken feather class.

Cotton foreign fiber fuzzy classification system and the method thereof that detects based on automatic vision provided by the present invention, by selecting color characteristic, shape facility and textural characteristics to constitute many proper vectors, and make up corresponding fuzzy classification device respectively, realization is to the kind identification of foreign fiber, thereby it is convenient, fast, measure content of profiled fibre in the unginned cotton exactly, the unginned cotton of give the selling deciding grade and level that grades is so that impel the conscious content that reduces foreign fiber in the cotton of cotton sellers.

Description of drawings

Fig. 1 the present invention is based on the cotton foreign fiber fuzzy classification method flow synoptic diagram that automatic vision detects;

Fig. 2 the present invention is based on the cotton foreign fiber fuzzy classification system structural representation that automatic vision detects.

Embodiment

Following examples are used to illustrate the present invention, but are not used for limiting the scope of the invention.

As shown in Figure 1, a kind of cotton foreign fiber fuzzy classification method provided by the present invention based on AVI, utilize the pixel coordinate information of the foreign fiber target that image segmentation obtains, together with original foreign fiber coloured image, extract the color characteristic of foreign fiber target, comprise the average of redness, green, blueness, brightness and the standard deviation of three components of RGB (RGB represents the color of three passages of red, green, blue); Extract the shape facility of foreign fiber target, comprise form factor, expansion ratio and Euler's numbers; Extract the textural characteristics of foreign fiber target circumscribed rectangular region, comprise standard deviation and contrast.Carry out preliminary classification by elementary fuzzy classification module according to these features then, wherein the elementary fuzzy classification device based on color characteristic utilizes color characteristic to judge tentatively which kind of foreign fiber foreign fiber to be identified may belong to, and obtains a degree of membership color level component; Elementary fuzzy classification device based on shape facility utilizes shape facility to judge tentatively which kind of foreign fiber foreign fiber to be identified may belong to, and obtains a degree of membership shape level component; Elementary fuzzy classification device based on textural characteristics utilizes textural characteristics to judge tentatively which kind of foreign fiber is foreign fiber target to be identified may belong to, and obtains a degree of membership texture stages component.These degree of membership components input to comprehensive degree of membership computing module, calculate the comprehensive degree of membership that foreign fiber target to be identified belongs to Polypropylence Sheet, cloth, the rope made of hemp, hair, polypropylene fibre silk and chicken feather.By the foreign fiber fuzzy sort module resulting six comprehensive degrees of membership are asked for maximal value at last, the classification with maximum comprehensive degree of membership is exactly the maximum possible classification of target to be identified.

As shown in Figure 2, a kind of cotton foreign fiber fuzzy classification system based on AVI provided by the present invention comprises characteristic extracting module, elementary fuzzy classification module, comprehensive degree of membership computing module and foreign fiber fuzzy sort module.

At first be defined as follows symbol:

I _RGB-based on the foreign fiber original color image of rgb color space;

I _RThe red component of-foreign fiber original color image;

I _GThe green component of-foreign fiber original color image;

I _BThe blue component of-foreign fiber original color image;

The equal value matrix of three color components of IM-foreign fiber original color image;

BW-foreign fiber the original color image is carried out the bianry image that Flame Image Process obtains;

The zone of certain foreign fiber target that FO-image segmentation obtains;

N _FoThe pixel count of certain foreign fiber target that-image segmentation obtains;

The circumscribed rectangular region of certain foreign fiber target that FB-image segmentation obtains;

N _FbPixel count in certain foreign fiber target boundary rectangle that-image segmentation obtains.

1, characteristic extracting module

Characteristic extracting module is used to carry out the extraction of color characteristic, shape facility and textural characteristics; Wherein color characteristic comprises the average of redness, green, blueness, brightness and the standard deviation of three components of RGB; Shape facility comprises form factor, expansion ratio and Euler's numbers; Textural characteristics comprises standard deviation and contrast.When carrying out feature extraction, what color characteristic and shape facility were got is the feature of target area FO, is the feature of target circumscribed rectangular region FB and textural characteristics is got.The color characteristic data are taken from original color image I _RGBCharacter shape data is taken from the foreign fiber original color image is carried out the bianry image BW that Flame Image Process obtains, the equal value matrix IM that the textural characteristics data are taken from three color components of foreign fiber original color image (rounds this matrix, just obtain a width of cloth gray level image, gray level from 0 to 255).Be the computing formula of using when carrying out feature extraction below.

The red average of foreign fiber target to be identified is defined as:

$\stackrel{\‾}{R}=\frac{1}{N_{\mathrm{fo}}}\underset{(i,j)\∈\mathrm{FO}}{\mathrm{\Σ}}{I}_R(i,j)$

The green average of foreign fiber target to be identified is defined as:

$\stackrel{\‾}{G}=\frac{1}{N_{\mathrm{fo}}}\underset{(i,j)\∈\mathrm{FO}}{\mathrm{\Σ}}{I}_G(i,j)$

The blue average of foreign fiber target to be identified is defined as:

$\stackrel{\‾}{B}=\frac{1}{N_{\mathrm{fo}}}\underset{(i,j)\∈\mathrm{FO}}{\mathrm{\Σ}}{I}_B(i,j)$

The brightness average of foreign fiber target to be identified is defined as:

$\stackrel{\‾}{I}=\frac{1}{N_{\mathrm{fo}}}\underset{(i,j)\∈\mathrm{FO}}{\mathrm{\Σ}}\mathrm{IM}(i,j)$

The color standard difference of foreign fiber target to be identified is defined as:

${\mathrm{\σ}}_{\mathrm{RGB}}=\sqrt{\frac{1}{N_{\mathrm{fo}}}\underset{(i,j)\∈\mathrm{FO}}{\mathrm{\Σ}}{(I_R(i,j)-\mathrm{IM}(i,j))}^2+{(I_G(i,j)-\mathrm{IM}(i,j))}^2+{(I_B(i,j)-\mathrm{IM}(i,j))}^2}$

The form factor of foreign fiber target to be identified is defined as:

S＝P ²/4πA

Wherein, P is the girth of foreign fiber target for this reason, i.e. the sum of all pixels of object boundary; A is the area of foreign fiber target for this reason, i.e. the sum of all pixels of target area.

The expansion ratio of foreign fiber target to be identified is defined as:

Ex＝A/A _BoundingBox

Wherein, A _BoundingBoxThe area of the boundary rectangle of foreign fiber target is represented with the sum of all pixels of this external rectangle area surrounded for this reason.The expansion ratio reflects the circularity of target area to a certain extent.

The Euler's numbers of foreign fiber target to be identified are defined as:

Eu＝N _components-N _holes

Wherein, N _ComponentsNumber for connected component in the foreign fiber target area; N _HolesNumber for hole in the foreign fiber target area.

The standard deviation of foreign fiber target to be identified is defined as:

${\mathrm{\σ}}_{\mathrm{bb}}=\sqrt{\underset{(i,j)\∈\mathrm{FB}}{\mathrm{\Σ}}{(\mathrm{IM}(i,j)-\stackrel{\‾}{I})}^{2}p\left(\mathrm{IM}(i,j)\right)}$

Wherein, (IM (i, j)) is that (promptly (i, j) number of times of Chu Xianing is divided by the total pixel number in zone for IM for i, the j) frequency that occurs in the circumscribed rectangular region of foreign fiber target for IM to p.Can obtain p (IM (i, j)) by the circumscribed rectangular region of foreign fiber target being carried out statistics of histogram.Standard deviation is used for reflecting the light and shade contrast degree of foreign fiber target and background.

The contrast of foreign fiber target to be identified is defined as:

$W_c=\underset{(i_1,j_1)\∈\mathrm{FB}}{\mathrm{\Σ}}\underset{(i_2,j_2)\∈\mathrm{FB}}{\mathrm{\Σ}}|\mathrm{IM}(i_1,j_1)-\mathrm{IM}(i_2,j_2)|p(\mathrm{IM}(i_1,j_1),\mathrm{IM}(i_2,j_2))$

Utilize the equal value matrix IM of color,, can obtain p (IM (i by calculating the gray level co-occurrence matrixes of foreign fiber target circumscribed rectangular region ₁, j ₁), IM (i ₂, j ₂)).Can reflect light and shade contrast between the neighborhood pixels based on the contrast of gray level co-occurrence matrixes, promptly pixel intensity is in the situation of change in space.

2, elementary fuzzy classification module

Described elementary fuzzy classification module is made up of three elementary fuzzy classification devices, is based on the elementary fuzzy classification device of color characteristic respectively, based on the elementary fuzzy classification device of shape facility with based on the elementary fuzzy classification device of textural characteristics; Obtain degree of membership color level component, shape level component and texture stages component that the foreign fiber target belongs to a different category respectively.

Elementary fuzzy classification device based on color characteristic can be defined as formally:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Color}}=f(\stackrel{\‾}{R},\stackrel{\‾}{G},\stackrel{\‾}{B},\stackrel{\‾}{I},{\mathrm{\σ}}_{\mathrm{RGB}})$

In this elementary fuzzy classification device based on color characteristic, parameters R, G, B, σ _RGBBe mainly used in the color of judging specific foreign fiber.Such as, hair, chicken feather should be (annotate: the method for the invention can't be discerned the hair and the chicken feather of white) of black, the rope made of hemp should be the Chinese ephedra look, the polypropylene fibre silk should be red, cyan or white etc.Parameter I is mainly used in brightness (representing its reflectance spectrum) information of reflection foreign fiber, and is lower such as the brightness of the foreign fiber that is faint in color, and the brightness of brightly painted foreign fiber is higher.Like this, take all factors into consideration color and brightness, just can carry out elementary fuzzy classification foreign fiber to be identified.

Elementary fuzzy classification device based on shape facility can be defined as formally:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}=g(S,\mathrm{Ex},\mathrm{Eu})$

In this elementary fuzzy classification device based on shape facility, parameter S f is mainly used in the circularity of judging foreign fiber, and when target is a bowlder, Sf reaches minimum value 1; Sf is generally greater than 1 during for other shape.Observe and find, the foreign fiber of sheet, best as the circularity of Polypropylence Sheet, the scraps of paper etc.; Velvet-like foreign fiber, relatively poor as the circularity of cloth, the rope made of hemp etc.; And the foreign fiber of wire is the poorest as the circularity of hair etc.Parameter Ex is mainly used in the degree of judging that foreign fiber is stretched, also reflects the circularity of target area to a certain extent.Generally speaking, the expansion ratio of Polypropylence Sheet is less, the expansion large percentage of the rope made of hemp, cloth and chicken feather, the expansion ratio maximum of hair and polypropylene fibre silk.Eu reflection be what of hole in the foreign fiber target area.Generally speaking, wire foreign fibers such as hair generally do not contain the hole, unless owing to reason such as curl, pile up produces the hole; The different fibre of sheets such as Polypropylence Sheet generally also has only less hole to occur, and large stretch of rope made of hemp, chicken feather, cloth, polypropylene fibre silk etc. are easy to generate more hole, but single rope made of hemp silk, chicken feather silk, polypropylene fibre silk generally the hole can not occur yet.

Elementary fuzzy classification device based on textural characteristics can be defined as formally:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}=h({\mathrm{\σ}}_{\mathrm{bb}},W_c)$

In this elementary fuzzy classification device based on textural characteristics, parameter σ _BbBe mainly used in the average contrast of the circumscribed rectangular region of judging the foreign fiber place, reflection be difference between grey scale pixel value and the area grayscale average, the while is also reflected the distribution situation of pixel on different grey-scale.Generally speaking, brighter because cotton is white in color, so the standard deviation of the gloomy different fibre of color is bigger, and the standard deviation of brightly painted different fibre is less.Parameter W _cThe grey scale change situation that also reflects foreign fiber place circumscribed rectangular region from another angle.

3, comprehensive degree of membership computing module

Described comprehensive degree of membership computing module is made up of six comprehensive degree of membership counters, is respectively the counter that is used to calculate plastics dry goods, cloth class, rope made of hemp class, hair class, polypropylene fibre silk class and the comprehensive degree of membership of chicken feather class; By described six counters, can obtain the comprehensive degree of membership that the foreign fiber target belongs to Polypropylence Sheet, cloth, the rope made of hemp, hair, polypropylene fibre silk and chicken feather respectively.

The computation model of comprehensive degree of membership is expressed as:

$\stackrel{\→}{\mathrm{\μ}}={\stackrel{\→}{w}}^{\mathrm{Color}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Color}}+{\stackrel{\→}{w}}^{\mathrm{Shape}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}+{\stackrel{\→}{w}}^{\mathrm{Texture}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}$

Wherein,

Be the weight of degree of membership color level component,

Be the weight of degree of membership shape level component,

Weight for degree of membership texture stages component; Why distributing different weights for different degree of membership components, is because of the classification contribution rate difference of different features to different types of foreign fiber.For example, confirm that according to the color characteristic of target to be identified this foreign fiber is a black, then this foreign fiber may be hair, chicken feather, also may be Polypropylence Sheet, and cloth can not be the rope made of hemp generally, can never be the polypropylene fibre silk.And for example, confirm that according to the shape facility of target to be identified this foreign fiber is elongated, then this foreign fiber may be hair, polypropylene fibre silk, also may be rope made of hemp silk or cloth silk, generally can not be Polypropylence Sheet; Consider color characteristic simultaneously, the elongated foreign fiber of black is hair most likely, also may be the chicken feather silk, also may be black cloth silk, but can assert substantially and be not the polypropylene fibre silk, because there is not black polypropylene fibre silk.Therefore, the weight of each component is comprehensively to determine according to the characteristics of foreign fiber.

4, foreign fiber fuzzy sort module

Described foreign fiber fuzzy sort module is used for resulting comprehensive degree of membership vector is asked for largest component, and the classification with maximum comprehensive degree of membership is exactly the maximum possible classification of foreign fiber target to be identified.Fuzzy classification model representation based on the comprehensive degree of membership of maximum is:

$\mathrm{FFClass}=\arg \max \left(\stackrel{\→}{\mathrm{\μ}}\right|\stackrel{\→}{\mathrm{\μ}}=\left[{\mathrm{\μ}}_A{\mathrm{\μ}}_B{\mathrm{\μ}}_C{\mathrm{\μ}}_D{\mathrm{\μ}}_E{\mathrm{\μ}}_F\right])$

For example, if in the comprehensive degree of membership vector of foreign fiber target to be identified, μ _AThe component maximum, then the classification of this foreign fiber maximum possible is a Polypropylence Sheet; If μ _DThe component maximum, then the classification of this foreign fiber maximum possible is exactly a hair.

Above embodiment only is used to illustrate the present invention; and be not limitation of the present invention; the those of ordinary skill in relevant technologies field; under the situation that does not break away from the spirit and scope of the present invention; can also make various variations and modification; therefore all technical schemes that are equal to also belong to category of the present invention, and scope of patent protection of the present invention should be defined by the claims.

Claims (12)

1, a kind of cotton foreign fiber fuzzy classification method that detects based on automatic vision is characterized in that, said method comprising the steps of:

S1 carries out feature selecting to color characteristic, shape facility and the textural characteristics of foreign fiber target to be identified, obtains three's proper vector separately;

S2, color feature vector, the shape facility vector sum texture feature vector of the foreign fiber target to be identified of extraction S1 gained;

S3, utilize resulting color feature vector among the S2, shape facility vector sum texture feature vector, obtain degree of membership color level component, degree of membership shape level component and degree of membership texture stages component that foreign fiber target to be identified belongs to different foreign fiber classifications;

S4 is weighted summation with described degree of membership color level component, degree of membership shape level component and degree of membership texture stages component, obtains the comprehensive degree of membership that target to be identified belongs to different foreign fiber classifications;

S5 belongs to the comprehensive degree of membership maximizing of different foreign fiber classifications to the foreign fiber target to be identified that obtains among the S4, and then the gained maximal value is maximum comprehensive degree of membership, and the classification with maximum comprehensive degree of membership is exactly the classification of foreign fiber target to be identified.

2, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 1, it is characterized in that, among the described step S1, described color feature vector comprises the average of redness, green, blueness and brightness and the standard deviation of three components of RGB; Described shape facility vector comprises form factor, expansion ratio and Euler's numbers; Described texture feature vector comprises standard deviation and contrast.

3, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 2, it is characterized in that, utilize resulting color feature vector among the S2, obtain the degree of membership color level component that foreign fiber target to be identified belongs to different foreign fiber classifications by following formula:

${\stackrel{\‾}{\mathrm{\μ}}}^{\mathrm{Color}}=f(\stackrel{\‾}{R},\stackrel{\‾}{G},\stackrel{\‾}{B},\stackrel{\‾}{I},{\mathrm{\σ}}_{\mathrm{RGB}})$

Wherein, R, G, B, I, σ _RGBRepresent redness, green, blueness and the brightness average of target to be identified and the standard deviation of three components of RGB respectively; F is the elementary fuzzy classification model of being set up based on color characteristic, The degree of membership color level component that belongs to different foreign fiber classifications for the foreign fiber target to be identified of model output.

4, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 2, it is characterized in that, utilize resulting shape facility vector among the S2, obtain the degree of membership shape level component that foreign fiber target to be identified belongs to different foreign fiber classifications by following formula:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}=g(S,\mathrm{Ex},\mathrm{Eu})$

Wherein, S, Ex, Eu represent form factor, expansion ratio and the Euler's numbers of target to be identified respectively; G is the elementary fuzzy classification model of being set up based on shape facility, The degree of membership shape level component that belongs to different foreign fiber classifications for the foreign fiber target to be identified of model output.

5, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 2, it is characterized in that, utilize resulting texture feature vector among the S2, utilize following formula to obtain the degree of membership texture stages component that foreign fiber target to be identified belongs to different foreign fiber classifications:

${\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}=h({\mathrm{\σ}}_{\mathrm{bb}},W_c)$

Wherein, σ _Bb, W _cStandard deviation and the contrast of representing target to be identified respectively; H is the elementary fuzzy classification model of being set up based on textural characteristics,

The degree of membership texture stages component that belongs to different foreign fiber classifications for the foreign fiber target to be identified of model output.

6, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 1 is characterized in that, among the described step S4, utilizes following formula to calculate the comprehensive degree of membership that foreign fiber target to be identified belongs to different foreign fiber classifications:

$\stackrel{\→}{\mathrm{\μ}}={\stackrel{\→}{w}}^{\mathrm{Color}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Color}}+{\stackrel{\→}{w}}^{\mathrm{Shape}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Shape}}+{\stackrel{\→}{w}}^{\mathrm{Texture}}{\stackrel{\→}{\mathrm{\μ}}}^{\mathrm{Texture}}$

Wherein,

Be the weight of degree of membership color level component,

Be the weight of degree of membership shape level component,

Weight for degree of membership texture stages component;

Be degree of membership color level component,

Be degree of membership shape level component,

Be degree of membership texture stages component;

Be the comprehensive degree of membership that foreign fiber target to be identified belongs to a different category.

7, the cotton foreign fiber fuzzy classification method that detects based on automatic vision as claimed in claim 1 is characterized in that, in step S5, utilizes following formula foreign fiber target to be identified to be belonged to the comprehensive degree of membership maximizing of different foreign fiber classifications:

$\mathrm{FFClass}=\arg \max \left(\stackrel{\→}{\mathrm{\μ}}\right|\stackrel{\→}{\mathrm{\μ}}=\left[{\mathrm{\μ}}_A{\text{\μ}}_B{\mathrm{\μ}}_C{\mathrm{\μ}}_D{\mathrm{\μ}}_E{\mathrm{\μ}}_F\right])$

Wherein,

The comprehensive degree of membership vector that belongs to a different category for foreign fiber target to be identified, arg represents to get the class code of max function gained maximal value representative but not maximal value itself, and FFClass is the class code of the foreign fiber that has maximum comprehensive degree of membership in the vector.

8, a kind of cotton foreign fiber fuzzy classification system that detects based on automatic vision is characterized in that described system comprises:

Characteristic extracting module is used to carry out the extraction of color characteristic, shape facility and textural characteristics;

Elementary fuzzy classification module links to each other with characteristic extracting module, is used to obtain degree of membership color level component, shape level component and the texture stages component that foreign fiber target to be identified belongs to a different category;

Comprehensive degree of membership computing module links to each other with elementary fuzzy classification module, is used to calculate the comprehensive degree of membership that foreign fiber target to be identified belongs to a different category;

The foreign fiber fuzzy sort module links to each other with comprehensive degree of membership computing module, is used for that resulting foreign fiber target to be identified is belonged to different classes of comprehensive degree of membership and asks for maximal value, and the classification with maximum comprehensive degree of membership is exactly the classification of target to be identified.

9, the cotton foreign fiber fuzzy classification system that detects based on automatic vision as claimed in claim 8 is characterized in that described color characteristic comprises the average of redness, green, blueness, brightness and the standard deviation of three components of RGB; Shape facility comprises form factor, expansion ratio and Euler's numbers; Textural characteristics comprises standard deviation and contrast.

10, the cotton foreign fiber fuzzy classification system that detects based on automatic vision as claimed in claim 8, it is characterized in that described elementary fuzzy classification module comprises elementary fuzzy classification device based on color characteristic, based on the elementary fuzzy classification device of shape facility with based on the elementary fuzzy classification device of textural characteristics.

11, the cotton foreign fiber fuzzy classification system that detects based on automatic vision as claimed in claim 8 is characterized in that the classification of described foreign fiber comprises plastics dry goods, cloth class, rope made of hemp class, hair class, polypropylene fibre silk and chicken feather class.

12, the cotton foreign fiber fuzzy classification system that detects based on automatic vision as claimed in claim 8, it is characterized in that, described comprehensive degree of membership computing module, comprise six comprehensive degree of membership counters, be respectively applied for the comprehensive degree of membership of calculating plastics dry goods, cloth class, rope made of hemp class, hair class, polypropylene fibre silk class and chicken feather class.

Images