CN101034440A - Identification method for spherical fruit and vegetables - Google Patents

Abstract

The invention is a recognition method of spherical fruit and vegetable category, transforming image obtained to the 2 r-g-b color model and LCD color model, according to classifier principle, discriminant F1, F2, F3, F4 were constructed separately under the two coordinate system, getting separate beeline omitted from the leaves and branches, the input image is divided into a small piece of the same size; orderly select two small plots B1 and B2, calculating the four direction of the gray co-occurrence matrix and two eigenvalues: entropy ENT and energy ASM, and the mean ENT and ASM; two adjacent plots of B1 and B2, such as discriminants F1, F2 , F3 and F4 is greater than zero, or the difference between ENT and ASM of B1 and B2 is less than the threshold T set, retaining B1, recognized as fruit; if the texture margin is greater than the threshold T and discriminants F1, F2, F3 and F4 is less than zero, discarding the entire plot B1; orderly take the next another plot as B2, repeating the steps until all plots have been disposed of. The invention can effectively recognize spherical fruits and vegetables, is simple to calculate with identify of high accuracy.

Description

The recognition methods of one kind spheral fruit

(1) technical field

The present invention relates to the recognition methods of a kind spheral fruit.

(2) background technology

Gathering of fruit and vegetables is a labor-intensive job.What fruits and vegetables were gathered is seasonal strong, and working environment is arduous, and labour intensity is big.The basic manpower that relies on of gathering of present stage, the efficiency ratio of gathering is lower, and human cost has accounted for sizable ratio in whole fruits and vegetables production cost.Therefore, realize that the robotization that fruits and vegetables are gathered has actual application value.

So far, domestic also do not have a fruit and vegetable picking robot that really puts into production use, and one of the main reasons is that the vision system technical barrier solves as yet fully.Usually, the working method of picking robot vision system: at first obtain the digitized image of fruits and vegetables, and then discern and determine the position of fruits and vegetables in the image by the associated picture Processing Algorithm.Sensor is the most important parts of Vision Builder for Automated Inspection, mainly comprises imageing sensor and range sensor etc.Imageing sensor can be CCD black and white camera, colour TV camera or stereo camera, generally is installed on mechanical arm or the end effector.Range sensor has laser ranging, ultrasonic, wireless and infrared sensor etc.

More early stage fruit and vegetable picking robot vision system is based on two-dimentional system mostly, and when the color contrast of leaf and fruits and vegetables was obvious, this two-dimentional system can successfully detect fruits and vegetables from leaf; Yet,, be difficult to discern if when having overlapping or its color of a plurality of fruits and vegetables to be similar to background colour.At this moment, generally be to detect fruits and vegetables according to the different spectral reflection characteristic of fruit and vegetable surfaces, after detecting fruits and vegetables, utilize three-dimensional visual sensor again fruits and vegetables accurately to be discerned, this is the most frequently used spectral reflectivity method, but under the natural lighting situation, owing to there are various interfere informations such as noise in the image, this method is often unsatisfactory.There is the researchist to utilize the shape of fruits and vegetables to discern and locate fruit, adopts the shape localization mode, generally require target to have complete boundary condition,, just be difficult to accomplish if object is blocked.The someone proposes a kind of new method of the Hough of being called conversion again, and its feature is not need whole profile information, locate the center of fruits and vegetables according to the curvature information of target fruits and vegetables shape, but this method is very consuming time.In a word, because the high complexity of environment, present picking robot vision system is under rule environment, and is relatively good as effect under the known illumination background, but at natural environmental condition, next as natural lighting is not very successful.So summary is got up, the method for current identification fruits and vegetables target, the main conversion of still passing through the color space model utilizes the color characteristic parameter to extract the fruits and vegetables target, and the identification problem that can solve is very limited.

(3) summary of the invention

In order to overcome the recognition methods of existing class spheral fruit effectively recognition category spheral fruit, calculation of complex, deficiency that accuracy of identification is low, the recognition methods of the invention provides a kind of effectively recognition category spheral fruit, calculating the kind spheral fruit simple, that accuracy of identification is high.

The technical solution adopted for the present invention to solve the technical problems is:

The recognition methods of one kind spheral fruit, this method may further comprise the steps:

(1), obtains fruits and vegetables image under the natural scene;

(2), the image that obtains is transformed to 2r-g-b color model and LCD color model simultaneously, the 2r-g-b color model is set up the 2r-g two-dimensional coordinate system, the LCD color model is set up the Y-Cr two-dimensional coordinate system;

(3), according to the sorter principle, to Y-C _rCharacteristic attribute Y under the two-dimensional coordinate system, C _rConstruct discriminant F respectively ₁, F ₂, set characteristic attribute value Y and C _rThe average vector m of fruit target _Fruit, leaf average vector m _Leaf, limb average vector m _Branch, its formula is: (1), (2):

F ₁＝[Y，C _r] ^T(m _fruit-m _leaf)-1/2[(m _fruit ^Tm _fruit)-(m _leaf ^Tm _leaf)] (1)

F ₂＝[Y，C _r] ^T(m _fruit-m _branch)-1/2[(m _fruit ^Tm _fruit)-(m _branch ^Tm _lbranch)] (2)

Characteristic attribute 2r and g to the 2r-g two-dimensional coordinate system construct discriminant F respectively ₃, F ₄, the average vector m ' f of the fruit target of setting characteristic attribute value 2r and g _Ruit, leaf average vector m ' _Leaf, limb average vector m ' _Branch, its formula is: (3), (4):

F ₃＝[2r，g] ^T(m′ _fruit-m′ _leaf)-1/2[(m′ _fruit ^Tm′ _fruit)-(m′ _leaf ^Tm′ _leaf)] (3)

F ₄＝[2r，g] ^T(m′ _fruit-m′ _branch)-1/2[(m′ _fruit ^Tm′ _fruit)-(m′ _branch ^Tm′ _lbranchz)] (4)

Obtain saving the straight line that separates of leaf and limb according to discriminant, and input picture is divided into equal-sized fritter, every block size is L * L, and L is an odd number;

(4), two fritter B of select progressively ₁And B ₂, calculate the gray level co-occurrence matrixes of its 4 directions respectively, an area size of establishing image is N _c* N _rPixel, and to establish gray level be G=0,1 ..., N _q-1, co-occurrence matrix P _{(d, q)}Be that a size is N _q* N _qSquare formation, comprise that all spacings are d, direction is q, and gray level is the frequency of pixel to occurring of a and b, P _{(d, q)}In element representation be P _{(a, b|d, q)}, optional two pixels in the zone (k, l) with (m, n), k wherein, m=1,2 ..., N _cN=1,2 ..., N _r

Calculate two eigenwerts with each gray level co-occurrence matrixes: entropy ENT and energy ASM, its formula are respectively (5), (6):

$\mathrm{ENT}=-\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}P(a,b)\log P(a,b)---\left(5\right)$

$\mathrm{ASM}=\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}{\left(P(a,b)\right)}^2---\left(6\right)$

Wherein, a, b is the gray level of remarked pixel respectively, p (a, b) expression gray level co-occurrence matrixes; And the eigenwert of calculating the gray level co-occurrence matrixes of average 4 directions obtains mean eigenvalue ENT and ASM;

(5), for two adjacent fritter B ₁And B ₂, as discriminant F ₁, F ₂, F ₃And F ₄Greater than 0 or B ₁And B ₂ENT and the difference of ASM less than preset threshold T, then keep B ₁, this fritter is confirmed as fruit; As the texture difference greater than threshold value T and discriminant F ₁, F ₂, F ₃And F ₄Less than 0, then abandon whole fritter B ₁And with B ₂Middle correlation parameter is composed to B ₁, then with B ₂Parameter be changed to sky, the order get another fritter as B backward ₂, repeating said steps disposes up to all fritters.

As preferred a kind of scheme: in described (5), for last fritter, if discriminant F ₁, F ₂, F ₃And F ₄Greater than 0, can be defaulted as fruit and directly keep, otherwise think background, directly abandon.

Technical conceive of the present invention is: because fruit has cluster at color space, this programme is used for reference mean distance sorter principle the fruits and vegetables image under the natural scene is carried out Classification and Identification.

The mean distance sorter is sorted out view data by analyzing different fruits and vegetables image digitization features.Sorting algorithm comprises training stage and test phase.In the training stage, distinguish the characteristic attribute of specific image, based on these attributes, produce the unique method of describing certain kinds, promptly produce training set, structure discriminant.Generally according to characteristic attribute m and n, utilize minimum distance classifier or mean distance sorter or other sorters can produce discriminant: F (m, n)=am+bn+c.A, b, c is an arbitrary constant, (m n)=0 gets final product as long as satisfy F.At test phase, utilize these feature spaces, i.e. these discriminants, each discriminant can be divided into image two parts, reaches the purpose of recognition image.

, construct 3 discriminants respectively for each color space and realize separating: F at each color space design category device at fruit target, leaf and limb _(fruit/leaf), be abbreviated as F ₁, be used for separating fruit and leaf; F _{(fruit/branch)}, be abbreviated as F ₂, be used for separating fruit and limb; F _{(leaf/branch)}, be abbreviated as F ₃, be used for separating leaf and limb.Because leaf and limb are considered background, so only need structure F ₁And F ₂Two discriminants.

Merge the characteristics of 2r-g-b color model and LCD (luminance and color difference) color model, obtain a kind of new color model LNM (LCD color model combinedwith Normalized-RGB color model).This model has been avoided the susceptibility of LCD color model to light, has avoided the 2r-g-b color model can't discern the drawback of the weak fruit of red color component value simultaneously, and the corresponding relation of LNM and RGB color model is shown below:

$\left\{\begin{array}{c}Y=0.299R+0.587G+0.114B\\ C_r=R-Y\\ 2r=2R/(R+G+B)\\ g=G/(R+G+B)\end{array}\right.$

Because merged two kinds of color spaces, in the process that realizes identification, we make up 4 discriminants altogether.

Simultaneously, adopt gray level co-occurrence matrixes to extract image texture characteristic.At first provide the notion of co-occurrence matrix: an area size of establishing image is N _c* N _rPixel, and to establish gray level be G=0,1 ..., N _q-1.Co-occurrence matrix P so _{(d, q)}Be that a size is N _q* N _qSquare formation, comprised that all spacings are d, direction is q, and gray level is the frequency of pixel to occurring of a and b.P _{(d, q)}In element can be expressed as P _{(a, b|d, q)}Optional two pixels in the zone (k, l) with (m, n), k wherein, m=1,2 ..., N _cN=1,2 ..., N _rThen:

P(a，b|d，q)＝∑[(k，l)，(m，n)]

Wherein, if

|k-m|＝q，|l-n|＝d，g(k，l)＝a，g(m，n)＝b

Set up, so

[(k，l)，(m，n)]＝1

Otherwise

[(k，l)，(m，n)]＝0

Wherein (k, (k l) locates the gray level of pixel to function g in l) representative.In addition, direction q can value be 0 °, 45 °, 90 °, 135 °.So for one given apart from d, can corresponding four gray level co-occurrence matrixes:

p(a，b|d，0°)＝∑[(k，l)，(m，n)]，k-m＝0，|l-n|＝d，(k，l)＝a，(m，n)＝b

p(a，b|d，45°)＝∑[(k，l)，(m，n)]，k-m＝d，l-n＝-d，(k，l)＝a，(m，n)＝b

Or k-m=-d, l-n=d, (k, l)=a, (m, n)=b

p(a，b|d，90°)＝∑[(k，l)，(m，n)]，|k-m|＝d，l-n＝0，(k，l)＝a，(m，n)＝b

p(a，b|d，135°)＝∑[(k，l)，(m，n)]，k-m＝d，l-n＝d，(k，l)＝a，(m，n)＝b

Or k-m=-d, l-n=d, (k, l)=a, (m, n)=b

Sample the respectively fruit and the leaf of multiple fruits and vegetables adopt gray level co-occurrence matrixes to extract characteristics of image, obtain distinguishing the textural characteristics value entropy and the energy of fruit and leaf, and its expression formula is respectively (5), (6):

$\mathrm{ENT}=-\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}P(a,b)\log P(a,b)---\left(5\right)$

$\mathrm{ASM}=\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}{\left(P(a,b)\right)}^2---\left(6\right)$

Wherein, a, b is the gray level of remarked pixel respectively, p (a, b) expression gray level co-occurrence matrixes.

Utilize region-growing method to solve the shortcoming of utilizing the solid color parameter less divided to occur preferably then based on sub-piece.

Beneficial effect of the present invention mainly shows: effectively the recognition category spheral fruit, calculate simple, accuracy of identification is high.

(4) description of drawings

Fig. 1 is schematic flow sheet of the present invention;

Fig. 2 is the sorter synoptic diagram;

Fig. 3 is a 2r-g-b sorter synoptic diagram;

Fig. 4 is a LCD sorter synoptic diagram;

Fig. 5 is the property distribution figure of fruit and background;

Fig. 6 is cut apart rectilinear according to what discriminant drew.

(5) embodiment

Below in conjunction with accompanying drawing the present invention is further described.

With reference to Fig. 1～Fig. 6, the recognition methods of a kind spheral fruit, this method may further comprise the steps: (1), obtain the fruits and vegetables image under the natural scene; (2), the image that obtains is transformed to 2r-g-b color model and LCD color model simultaneously, the 2r-g-b color model is set up the 2r-g two-dimensional coordinate system, the LCD color model is set up Y-C _rTwo-dimensional coordinate system; (3), according to the sorter principle, to Y-C _rCharacteristic attribute Y under the two-dimensional coordinate system, C _rConstruct discriminant F respectively ₁, F ₂, set characteristic attribute value Y and C _rThe average vector m of fruit target _Fruit, leaf average vector m _Leaf, limb average vector m _Branch, its formula is: (1), (2):

F ₁＝[Y，C _r] ^T(m _fruit-m _leaf)-1/2[(m _fruit ^Tm _fruit)-(m _leaf ^Tm _leaf)] (1)

F ₂＝[Y，C _r] ^T(m _fruit-m _branch)-1/2[(m _fruit ^Tm _fruit)-(m _branch ^Tm _lbranch)] (2)

Characteristic attribute 2r and g to the 2r-g two-dimensional coordinate system construct discriminant F respectively ₃, F ₄, the average vector m ' of the fruit target of setting characteristic attribute value 2r and g _Fruit, leaf average vector m ' _Leaf, limb average vector m ' _Branch, its formula is: (3), (4):

F ₃＝[2r，g] ^T(m′ _fruit-m′ _leaf)-1/2[(m′ _fruit ^Tm′f _ruit)-(m′ _leaf ^Tm′ _leaf)] (3)

F ₄＝[2r，g]T(m′ _fruit-m′ _branch)-1/2[(m′ _fruit ^Tm′ _fruit)-(m′ _branch ^Tm′ _lbranch)] (4)

Obtain saving the straight line that separates of leaf and limb according to discriminant, and input picture is divided into equal-sized fritter, every block size is L * L, and L is an odd number;

(4), two fritter B of select progressively ₁And B ₂, calculate the gray level co-occurrence matrixes of its 4 directions respectively, an area size of establishing image is N _c* N _rPixel, and to establish gray level be G=0,1 ..., N _q-1, co-occurrence matrix P _{(d, q)}Be that a size is N _q* N _qSquare formation, comprise that all spacings are d, direction is q, and gray level is the frequency of pixel to occurring of a and b, P _{(d, q)}In element representation be P _{(a, b|d, q)}, optional two pixels in the zone (k, l) with (m, n), k wherein, m=1,2 ..., N _cN=1,2 ..., N _r

Calculate two eigenwerts with each gray level co-occurrence matrixes: entropy ENT and energy ASM, its formula are respectively (5), (6):

$\mathrm{ENT}=-\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}P(a,b)\log P(a,b)---\left(5\right)$

$\mathrm{ASM}=\underset{i=0}{\overset{N_q-1}{\mathrm{\Σ}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\Σ}}}{\left(P(a,b)\right)}^2---\left(6\right)$

Wherein, a, b is the gray level of remarked pixel respectively, p (a, b) expression gray level co-occurrence matrixes; And the eigenwert of calculating the gray level co-occurrence matrixes of average 4 directions obtains mean eigenvalue ENT and ASM;

(5), for two adjacent fritter B ₁And B ₂, as discriminant F ₁, F ₂, F ₃And F ₄Greater than 0 or B ₁And B ₂ENT and the difference of ASM less than preset threshold T, then keep B ₁, this fritter is confirmed as fruit; As the texture difference greater than threshold value T and discriminant F ₁, F ₂, F ₃And F ₄Less than 0, then abandon whole fritter B ₁And with B ₂Middle correlation parameter is composed to B ₁, then with B ₂Parameter be changed to sky, the order get another fritter as B backward ₂, repeating said steps disposes up to all fritters.

In described (5), for last fritter, if discriminant F ₁, F ₂, F ₃And F ₄Greater than 0, can be defaulted as fruit and directly keep, otherwise think background, directly abandon.

Concrete steps in the present embodiment are:

(1), obtains fruits and vegetables image under the natural scene by digital camera or camera.

(2), the image that obtains is transformed to 2r-g-b color model and LCD color model simultaneously.

(3), according to the sorter principle, to the characteristic attribute 2r under the 2r-g two-dimensional coordinate system of 2r-g-b model, the Y-C of g and LCD _rCharacteristic attribute Y under the two-dimensional coordinate system, C _rConstruct discriminant respectively, produce 4 discriminant F ₁, F ₂, F ₃, F ₄Simultaneously, input picture is divided into equal-sized fritter, every block size is L * L, and L is an odd number.

Illustrate:

Suppose each the training class represent by an average vector:

$m_j=1/N_j\underset{x\∈w_j}{\mathrm{\Σ}}x$ for j＝1，2，…，M

Here N _iBe class W _jThe quantity of training mode vector.Suppose to have based on characteristic attribute value Y and C _rFruit target, leaf and limb, suppose characteristic attribute is described in the two-dimensional feature space, as shown in Figure 5.And can obtain mean value is: m _Fruit=[94.14,114.26] ^T, m _Leaf=[77.02,218.45] ^T, m _Branch=[52.88,33.67] ^T, as representing with # among Fig. 5.

Based on this point, can be by judging near m _jPrototype, distribute given arbitrarily pattern x to give class.As adopt Euclidean distance as measuring similarity distance, the prototype distance is:

F _j(x)＝‖x-m _j‖ for j＝1，2，…，M

This is equivalent to:

F _j(x)＝x ^Tm _j-1/2(m _j ^Tm _j) for j＝1，2，…，M

In this example, the mean value of each fruit that obtains of simple hypothesis is [a, b], and the mean value of leaf is [c, d], and the mean value of branch is [e, f].Can obtain so

$\left\{\begin{array}{c}{F}_{\mathrm{fruit}}=\mathrm{aY}+\mathrm{bCr}-1/2(a^2+b^2)\\ F_{\mathrm{leaf}}=\mathrm{cY}+\mathrm{dCr}-1/2({\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="A20071006669500152.png"\; state="\{\{state\}\}">c</figure-callout>}}^2+d^2)\\ F_{\mathrm{branch}}=\mathrm{eY}+\mathrm{fCr}-1/2({\mathrm{<figure-callout\; id="2"\; label="e"\; filenames="A20071006669500152.png"\; state="\{\{state\}\}">e</figure-callout>}}^2+f^2)\end{array}\right.$

According to this method, can calculate following decision function:

$\left\{\begin{array}{c}{F}_{\mathrm{fruit}}=94.14Y+114.26\mathrm{Cr}-10958.84\\ F_{\mathrm{leaf}}=77.02Y+218.45\mathrm{Cr}-26826.24\\ F_{\mathrm{branch}}=52.88Y+33.67\mathrm{Cr}-1964.98\end{array}\right.$

At last, according to these decision functions, can separate class w _iAnd w _jDiscriminant satisfy:

F _i(x)-F _j(x)＝0

That is:

$\left\{\begin{array}{c}{F}_{\mathrm{fruit}/\mathrm{leaf}}=(a-c)Y+(b-d)\mathrm{Cr}+1/2(a^2+b^2-c^2-d^2)=0\\ F_{\mathrm{fruit}/\mathrm{branch}}=(a-e)Y+(b-f)\mathrm{Cr}+1/2(a^2+b^2-e^2-f^2)=0\\ F_{\mathrm{leaf}/\mathrm{branch}}=(c-e)Y+(d-f)\mathrm{Cr}+1/2({\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="A20071006669500152.png"\; state="\{\{state\}\}">c</figure-callout>}}^2+d^2-e^2-f^2)=0\end{array}\right.$

In this example, obtain discriminant F ₁, F ₂, F ₃As follows:

$\left\{\begin{array}{c}{F}_{\mathrm{fruit}/\mathrm{leaf}}=17.12Y-104.19\mathrm{Cr}+15867.4=0\\ F_{\mathrm{fruit}/\mathrm{branch}}=41.26Y+80.59\mathrm{Cr}+8993.86=0\\ F_{\mathrm{leaf}/\mathrm{branch}}=24.14Y+184.78\mathrm{Cr}+24861.26=0\end{array}\right.$

According to these discriminants, can draw the separation straight line, save the straight line that separates of leaf and limb, as shown in Figure 6.

If the different discriminant that obtains of the prototype that adopts distance also is different, as adopting standardized Euclidean distance and Ma Halangnuobisi distance etc.

(4), two fritter B of select progressively ₁And B ₂, calculate two eigenwerts of gray level co-occurrence matrixes He each gray level co-occurrence matrixes of its 4 directions respectively:: entropy ENT and energy ASM, its formula are respectively (5), (6):

$\mathrm{ENT}=-\underset{i=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}P(a,b)\log P(a,b)---\left(5\right)$

$\mathrm{ASM}=\underset{i=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}{\left(P(a,b)\right)}^2---\left(6\right)$

Wherein, a, b is the gray level of remarked pixel respectively, p (a, b) expression gray level co-occurrence matrixes;

(5), the eigenwert of the gray level co-occurrence matrixes of average 4 directions obtains mean eigenvalue ENT and ASM as the textural characteristics of differentiating.

(6), for two adjacent fritter B ₁And B ₂, according to the discriminant F of color model structure ₁, F ₂, F ₃, F ₄Judge with the textural characteristics value.If discriminant is all greater than 0 or B ₁And B ₂ENT and the difference of ASM less than preset threshold T, then keep B ₁, be pressed into a storehouse.Simultaneously with B ₂Middle correlation parameter is composed to B ₁, then with B ₂Parameter be changed to sky.Order is got another fritter as B backward ₂, repeat above-mentioned steps.If the texture difference is greater than threshold value T and discriminant F ₁, F ₂, F ₃, F ₄All, then abandon whole fritter B less than 0 ₁Dispose up to all fritters.

Claims (2)

1, the recognition methods of a kind spheral fruit, this method may further comprise the steps:

(1), obtains fruits and vegetables image under the natural scene;

(2), the image that obtains is transformed to 2r-g-b color model and LCD color model simultaneously, the 2r-g-b color model is set up the 2r-g two-dimensional coordinate system, the LCD color model is set up Y-C _rTwo-dimensional coordinate system;

(3), according to the sorter principle, to Y-C _rCharacteristic attribute Y under the two-dimensional coordinate system, C _rConstruct discriminant F respectively ₁, F ₂, set characteristic attribute value Y and C _rThe average vector m of fruit target _Fruit, leaf average vector m _Leaf, limb average vector m _Branch, its formula is: (1), (2):

F ₁＝[Y，C _r] ^T(m _fruit-m _leaf)-1/2[(m _fruit ^Tm _fruit)-(m _leaf ^Tm _leaf)] (1)

F ₂＝[Y，C _r] ^T(m _fruit-m _branch)-1/2[(m _fruit ^Tm _fruit)-(m _branch ^Tm _lbranch)] (2)

Characteristic attribute 2r and g to the 2r-g two-dimensional coordinate system construct discriminant F respectively ₃, F ₄, the average vector m ' of the fruit target of setting characteristic attribute value 2r and g _Fruit, leaf average vector m ' _Leaf, limb average vector m ' _Branch, its formula is: (3), (4):

F ₃＝[2r，g] ^T(m′ _fruit-m′ _leaf)-1/2[(m′ _fruit ^Tm′ _fruit)-(m′ _leaf ^Tm′ _leaf)] (3)

F ₄＝[2r，g] ^T(m′ _fruit-m′ _branch)-1/2[(m′ _fruit ^Tm′ _fruit)-m′ _branch ^Tm′ _lbranch)] (4)

Obtain saving the straight line that separates of leaf and limb according to discriminant, and input picture is divided into equal-sized fritter, every block size is L * L, and L is an odd number;

(4), two fritter B of select progressively ₁And B ₂, calculate the gray level co-occurrence matrixes of its 4 directions respectively, an area size of establishing image is N _c* N _rPixel, and to establish gray level be G=0,1 ..., N _q-1, co-occurrence matrix P _{(d, q)}Be that a size is N _q* N _qSquare formation, comprise that all spacings are d, direction is q, and gray level is the frequency of pixel to occurring of a and b, P _{(d, q)}In element representation be P _{(a, b|d, q)}, optional two pixels in the zone (k, l) with (m, n), k wherein, m=1,2 ..., N _cN=1,2 ..., N _r

Calculate two eigenwerts with each gray level co-occurrence matrixes: entropy ENT and energy ASM, its formula are respectively (5), (6):

$\mathrm{ENT}=-\underset{i=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}P(a,b)\log P(a,b)...\left(5\right)$

$\mathrm{ASM}={\underset{i=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{N_q-1}{\mathrm{\&Sigma;}}}\left(P(a,b)\right)}^2...\left(6\right)$

Wherein, a, b is the gray level of remarked pixel respectively, p (a, b) expression gray level co-occurrence matrixes; And the eigenwert of calculating the gray level co-occurrence matrixes of 4 directions obtains mean eigenvalue ENT and ASM;

(5), for two adjacent fritter B ₁And B ₂, as discriminant F ₁, F ₂, F ₃And F ₄Greater than 0 or B ₁And B ₂ENT and the difference of ASM less than preset threshold T, then keep B ₁, this fritter is confirmed as fruit; As the texture difference greater than threshold value T and discriminant F ₁, F ₂, F ₃And F ₄Less than 0, then abandon whole fritter B ₁And with B ₂Middle correlation parameter is composed to B ₁, then with B ₂Parameter be changed to sky, the order get another fritter as B backward ₂, repeating said steps disposes up to all fritters.

2, the recognition methods of a kind spheral fruit as claimed in claim 1 is characterized in that: in described (5), for last fritter, if discriminant F ₁, F ₂, F ₃And F ₄Greater than 0, can be defaulted as fruit and directly keep, otherwise think background, directly abandon.

Images