CN116797544B - Surface defect extraction method for fruit and vegetable post-harvest treatment equipment - Google Patents

Abstract

The invention discloses a surface defect extraction method for fruit and vegetable post-harvest processing equipment, which comprises the steps of obtaining three optimized thresholds for dividing four color card number sets by maximizing color card number three-threshold maximum inter-class variance weighted by comprehensive adjacent color card number average element difference mutual information quantity and maximum probability color card number average element difference mutual information quantity of four color card number sets of background, mature, immature and surface defects, and then comparing color card numbers corresponding to element combinations of all pixel points in an image with the three optimized thresholds to finish the division of the four color card number sets so as to realize the surface defect extraction for the fruit and vegetable post-harvest processing equipment.

Description

A surface defect extraction method for fruit and vegetable post-harvest processing equipment

Technical Field

The invention belongs to the technical field of post-harvest processing of fruits and vegetables, and specifically relates to maximizing the maximum inter-class variance of three threshold values of color card numbers weighted by comprehensive mutual information of average element differences of adjacent color card numbers and mutual information of average element differences of maximum probability color card numbers for four color card number sets of background, maturity, immaturity and surface defects, thereby obtaining three optimized threshold values for dividing the four color card number sets, thereby segmenting the background, maturity, immaturity and surface defect related areas in the fruit and vegetable color image given by post-harvest processing equipment of fruits and vegetables to complete the surface defect extraction work.

Background Art

When processing fruit and vegetable images during the post-harvest sorting process, people are usually only interested in certain parts of them. The areas where these points of interest are focused can be defined as target areas (also called foreground areas). In most cases, the target area has certain common characteristics, while the rest is the background area. Only by separating the target from the background and extracting its characteristics can it be identified and analyzed, and then the target can be applied at a deeper level. In pattern recognition and machine vision systems, the importance of image segmentation is unquestionable and is the basis for further understanding of images. Many important subsequent tasks in image processing, such as feature extraction, image analysis, pattern recognition, and image understanding, are based on image segmentation technology. Many scholars at home and abroad have proposed various segmentation algorithms. Among them, the threshold-based segmentation method has attracted much attention and love from researchers because of its advantages such as low threshold calculation difficulty and strong robustness. A Japanese researcher named Otsu Hideyuki proposed a new threshold segmentation method (referred to as Otsu method) based on the principle of maximum inter-class variance, which is recognized as a classic algorithm in image threshold segmentation technology. The Otsu method has the characteristics of small computational complexity and does not change dramatically due to changes in image brightness or contrast.

The LAB color model consists of three elements, one of which is brightness (L). A and B are two color channels, where A includes colors ranging from dark green (low brightness value) to gray (medium brightness value) to bright pink (high brightness value), and B includes colors ranging from bright blue (low brightness value) to gray (medium brightness value) to yellow (high brightness value).

The Pantone color card is an internationally used standard color card, covering the color communication system in the fields of printing, textiles, plastics, drawing, digital technology, etc., and has become the international unified standard language for exchanging color information today. Each PANTONE color has its own unique number. For example, the color number in the PANTONE printing color card is composed of 3 or 4 digits plus the letter C or U, such as pantone 100c or 100u, or pantone 1205C or 1205U. The letter C means the performance of this color on coated paper, and the letter U means the performance of this color on offset paper.

Considering the complexity and real-time nature of image processing in fruit and vegetable post-harvest processing equipment, the image information in practical applications contains a large amount of data such as ripe color, unripe color, and surface defects. Therefore, it takes a lot of time to select a set of scientific and reasonable thresholds from the threshold set. It can be seen that the design can be carried out at the specific application level of fruit and vegetable post-harvest processing equipment, which will be conducive to enriching and expanding the connotation of surface defect extraction methods.

Summary of the invention

The purpose of the present invention is to classify the element combination corresponding to the pixel points in the fruit and vegetable color image that needs to be extracted for surface defects given by the fruit and vegetable post-harvest processing equipment into four color card number sets of background, mature, immature, and surface defects, determine the maximum inter-class variance of the color card number three thresholds weighted by the average element difference mutual information of the comprehensive adjacent color card numbers and the average element difference mutual information of the maximum probability color card number for the four color card number sets, and then compare the color card number corresponding to the element combination of the pixel point with the three optimized thresholds for dividing the four color card number sets in step 2 to determine the corresponding color card number set number, complete the division of the four color card number sets, and thus realize the surface defect extraction for the fruit and vegetable post-harvest processing equipment;

To achieve the above purpose, the technical solution proposed by the present invention is as follows:

A surface defect extraction method for fruit and vegetable post-harvest processing equipment, the method comprising the following steps:

Step 1: For the fruit and vegetable post-harvest processing equipment, a color image of fruits and vegetables that needs to be extracted for surface defects is given. After the brightness of each pixel in the color image of fruits and vegetables is normalized according to a certain color model, a certain color card number corresponding to the three elements of the color model is output. The obtained color card number is set to be classified into four color card number sets: background, mature, immature, and surface defects. The probability of the four color card number sets and the corresponding element average values are given. The probability of different color card number sets being jointly generated and the corresponding element average values are given. The mutual information of the average element difference of adjacent color card numbers is given when the color card number difference between a certain color card number set and the next color card number set is within the specified conditions. The color card number when the probability of each pixel point in the image of a certain color card number set and another color card number set is the largest is given. The maximum inter-class variance of the three thresholds of the color card number weighted by the comprehensive mutual information of the average element difference of adjacent color card numbers and the mutual information of the average element difference of the maximum probability color card number for the four color card number sets is determined;

Step 2: Based on the historical threshold values of the three color card numbers as the initial values, the threshold combinations are continuously increased or decreased by the product of the color card number threshold changes and the cycle process count to be classified into the threshold permutation combination set and the calculation results are classified into the color card number three-threshold maximum inter-class variance value set. If the color card number three-threshold maximum inter-class variance value set does not add a larger value, then the threshold combination corresponding to the largest value in the color card number three-threshold maximum inter-class variance value set is the three optimized thresholds for dividing the four color card number sets;

Step 3: For the pixel points in the image, the color card number corresponding to the element combination of the pixel point can be compared with the three optimized thresholds for dividing the four color card number sets in step 2 to determine the corresponding color card number set number, and then the color card number set numbers corresponding to all the pixel points in the image are output to complete the division of the four color card number sets to extract the surface defects;

Furthermore, in step 1, the brightness of each pixel in the fruit and vegetable color image is normalized according to a certain color model, and then a certain color card number corresponding to the three elements of the color model is output. The color card number obtained can be divided into four color card number sets of background, ripe, immature, and surface defects. Specifically, the color card number corresponding to the element combination (l, a, b) after brightness normalization according to a certain color model in the fruit and vegetable color image can be obtained from arrive Among them, l, a, and b are the values of the three elements given by the color model for the pixel points in the fruit and vegetable color image. are the values of the three elements corresponding to the color with the smallest color card number in the fruit and vegetable color image, They are the values of the three elements corresponding to the color with the largest color card number in the fruit and vegetable color image, and are divided into 4 color card number sets, Ω ₁ , Ω ₂ , Ω ₃ and Ω _4. The 4 color card number sets can be characterized as background, ripe, immature and surface defects. The specific forms are as follows:

Ω ₂ ={Γ(x ₁ ,y ₁ ,z ₁ )+1,...,Γ(x ₂ ,y ₂ ,z ₂ )}

Ω ₃ ={Γ(x ₂ ,y ₂ ,z ₂ )+1,...,Γ(x ₃ ,y ₃ ,z ₃ )}

Wherein, Γ is a function that inputs a combination of elements and outputs a color card number, (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) and (x ₃ , y ₃ , z ₃ ) are element combinations corresponding to the color card numbers that divide the four color card number sets of Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ ;

Furthermore, in step 1, the probabilities of generating the four color card number sets and the corresponding element average values are specifically: the probabilities of generating the four color card number sets Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ are κ ₁ , κ ₂ , κ ₃ and κ ₄ respectively, and the specific form is as follows:

Among them, i∈[1,2,3,4], ε _(l,a,b) is the probability of the element combination (l,a,b) at the pixel point in the image;

The element average values corresponding to the four color card number sets Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ are Υ ₁ , Υ ₂ , Υ ₃ and Υ ₄ , respectively. The specific forms are as follows:

Among them, α, β and γ are the weight coefficients of factors L, A and B in the average value of factors respectively;

Further, in step 1, the probabilities of different sets of color card numbers being jointly generated and the corresponding element average values are specifically: the probability of Ω ₁ and Ω ₂ being jointly generated is κ _1,2 , the probability of Ω ₁ and Ω ₃ being jointly generated is κ _1,3 , the probability of Ω ₁ and Ω ₂ being jointly generated is κ _1,4 , the probability of Ω ₁ and Ω ₂ being jointly generated is κ _1,3 , the probability of Ω ₁ and Ω ₄ being jointly generated is κ _1,4 , the probability of Ω ₂ and Ω ₃ being jointly generated is κ _2,3 , the probability of Ω ₂ and Ω ₄ being jointly generated is κ _2,4 , and the probability of Ω ₃ and Ω ₄ being jointly generated is κ _3,4 , and the specific form is as follows:

Among them, i,j∈[1,2,3,4], i≠j and i＜j;

The average values of the elements corresponding to Ω _i and Ω _j are Ψ _i,j , where i,j∈[1,2,3,4], i≠j and i＜j. The specific form is as follows:

Further, in step 1, the mutual information of average element difference between adjacent color card numbers is given by a certain color card number set and the next color card number set when the color card number difference is within the limited condition: for Ω _i and Ω j satisfying i, j∈[1,2,3,4] and i+1＝ _j , the limited value of the difference between adjacent color card numbers is set to k, then any color card number Γ(x,y,z) in Ω _i and any color card number Γ(x′,y′,z′) in the adjacent Ω _j can be calculated within the limited condition Γ(x,y,z)-Γ(x′,y′,z′)≤k for the mutual information of average element difference between adjacent color card numbers ξ _i,j,k for Ω _i and Ω _j , and the specific form is as follows:

in, is any symbol;

Further, in step 1, the maximum probability color card number average element difference mutual information is given by the color card number when the probability of the pixel point in the image of a certain color card number set and another color card number set is the maximum: for Ω _i and Ω j satisfying i, j∈[1,2,3,4] and j＞i+1, the maximum probability color card number average element difference mutual information for Ω _i and Ω _j can be calculated by the color card number Γ(x _i ,y _i ,z _i ) when the probability of the pixel point in Ω _i is the maximum and the color card number Γ ₍ x _j ,y _j ,z _j ) when the probability of the pixel point in Ω _j is the maximum The specific form is as follows:

in,

Further, in step 1, the method of determining the weighted three-threshold maximum inter-class variance of the color card number of the comprehensive adjacent color card number average element difference mutual information and the maximum probability color card number average element difference mutual information for the _four color card number sets is specifically as follows: determining the adjacent color card number average element difference mutual information ξ _i,j,k and the maximum probability color card number average element difference mutual information for Ω ₁ , Ω ₂ , Ω ₃ and Ω 4 The maximum inter-class variance of the weighted three-threshold color card number is σ ² [Γ(x ₁ ,y ₁ ,z ₁ ),Γ(x ₂ ,y ₂ ,z ₂ ),Γ(x ₃ ,y ₃ ,z ₃ )], and the specific form is as follows:

Among them, Θ ₁ ={(1,2),(2,3),(3,4)}, Θ ₂ ={(1,3),(1,4),(2,4)};

Furthermore, step 2 specifically includes:

Step 2.1: Number the three color cards to the historical thresholds and As the initial values of the color card number thresholds Γ(x ₁ ,y ₁ ,z ₁ ), Γ(x ₂ ,y ₂ ,z ₂ ) and Γ(x ₃ ,y ₃ ,z ₃ ) in the current fruit and vegetable color image processing, the color card number threshold change is set to Δ, the loop process count t is 0, the threshold combination (Γ(x ₁ ,y ₁ ,z ₁ ),Γ(x ₂ ,y ₂ ,z ₂ ),Γ(x ₃ ,y ₃ ,z ₃ )) is included in the threshold permutation combination set ν(t=0), the threshold combination in ν(t=0) is calculated according to the specific form of the maximum inter-class variance of the three thresholds of the color card number described in step 1, and the calculation result is included in the maximum inter-class variance value set υ(t=0) of the three thresholds of the color card number;

Step 2.2: Add 1 to t instead of the original t, set the intermediate variable θ to Δ·t or 0, and classify the threshold combinations (Γ(x ₁ ,y ₁ ,z ₁ )±θ,Γ(x ₂ ,y ₂ ,z ₂ )±θ,Γ(x ₃ ,y ₃ ,z ₃ )±θ) into the threshold permutation combination set ν(t);

Step 2.3: Calculate the threshold combination in ν(t) according to the specific form of the maximum inter-class variance of the three thresholds of the color card number described in step 1, and include the calculation result in the maximum inter-class variance value set υ(t) of the three thresholds of the color card number;

Step 2.4: Compare υ(t) with υ(t-1) to see if there is a larger value. If υ(t) is larger than υ(t-1), go to step 2.2, otherwise go to step 2.5;

Step 2.5: The largest value in υ(t) corresponds to and The threshold combination is set to three optimized thresholds for dividing the four color card number sets of Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ ;

Furthermore, step 3 specifically includes: for a pixel point (m, n) in the image, the corresponding color card number set number f(m, n) can be determined by the element combination (l, a, b) corresponding to the pixel point (m, n), and the specific form is as follows:

The color card number set numbers corresponding to all pixels in the output image are divided into four color card number sets Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ to extract the color card number set corresponding to the surface defects.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a diagram showing the steps of surface defect extraction for fruit and vegetable post-harvest processing equipment;

FIG2 is an example of a sample set of navel orange color images;

FIG3 shows the surface defect area of the navel orange color image after being divided by the color card number set.

DETAILED DESCRIPTION

The present invention will be further described below in conjunction with the accompanying drawings by using a specific implementation of the surface defect extraction method for fruit and vegetable post-harvest processing equipment applied to navel orange color images.

As shown in FIG1 , a surface defect extraction method for fruit and vegetable post-harvest processing equipment according to an embodiment of the present invention comprises the following steps:

Step 1: For the navel orange color image that needs to be extracted for surface defects given by the fruit and vegetable post-harvest processing equipment, the LAB color model and PANTONE color card mode can be used, where the PANTONE color card number is the first 3 digits of PANTONE NC XXXC or XXXXC, where X represents a number. The PANTONE color card number corresponding to the element combination (l, a, b) in the navel orange color image after brightness normalization according to the LAB color model can be obtained from arrive Among them, l, a, and b are the values of the three elements given by the color model for the pixels in the navel orange color image. are the values of the three elements corresponding to the color with the smallest color card number in the navel orange color image, They are the values of the three elements corresponding to the color with the largest color card number in the navel orange color image, and are divided into 4 color card number sets, Ω ₁ , Ω ₂ , Ω ₃ and Ω _4. The 4 color card number sets can be characterized as background, mature, immature, and surface defects. The specific form is as follows:

Ω ₂ ={Γ(x ₁ ,y ₁ ,z ₁ )+1,...,Γ(x ₂ ,y ₂ ,z ₂ )}

Ω ₃ ={Γ(x ₂ ,y ₂ ,z ₂ )+1,...,Γ(x ₃ ,y ₃ ,z ₃ )}

Wherein, Γ is a function that inputs a combination of elements and outputs a color card number, (x ₁ , y ₁ , z ₁ ), (x ₂ , y ₂ , z ₂ ) and (x ₃ , y ₃ , z ₃ ) are element combinations corresponding to the color card numbers that divide the four color card number sets of Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ ;

The probability of Ω _i and Ω _j being generated together is κ _i,j , which is in the following form:

Among them, i,j∈[1,2,3,4], ε _(l,a,b) is the probability of the element combination in the pixel point of the image, and _κi is the probability of the element combination set _Ωi ;

The average value of the elements corresponding to Ω _i is Υ _i , which is in the following form:

The common element averages of Ω _i and Ω _j are Ψ _i,j , where i,j∈[1,2,3,4], i≠j and i＜j. The specific form is as follows:

For Ω _i and Ω j that satisfy i, j ∈ [1, 2, 3, 4] and i + 1 = _j , set the limit value of the difference between adjacent color card numbers to k, then the average element difference mutual information ξ i, j, k for adjacent color card numbers of Ω _i and Ω _j within the limit condition Γ (x, y, z) - Γ (x', y', z') ≤ k can be calculated through any color card number Γ (x, y, z) in Ω _i and any color card number Γ (x', _y' , z') in the adjacent Ω _j . The specific form is as follows:

in, is any symbol;

For Ω _i and Ω _j that satisfy i, j∈[1,2,3,4] and j＞i+1, the average element difference mutual information of the maximum probability color card numbers for Ω i and Ω _j can be calculated by the color card number Γ(x _i ,y _i , _zi ) when the probability of the pixel point in Ω _i is the largest and the color card number Γ ₍ x _j ,y _j ,z _j ) when the probability of the pixel point in Ω _j is the largest The specific form is as follows:

in,

For the adjacent color card numbers of Ω ₁ , Ω ₂ , Ω ₃ and Ω _4, the average element difference mutual information ξ _i,j,k and the maximum probability color card number average element difference mutual information The maximum inter-class variance of the weighted three-threshold color card number is σ ² [Γ(x ₁ ,y ₁ ,z ₁ ),Γ(x ₂ ,y ₂ ,z ₂ ),Γ(x ₃ ,y ₃ ,z ₃ )], and the specific form is as follows:

Among them, Θ ₁ ={(1,2),(2,3),(3,4)}, Θ ₂ ={(1,3),(1,4),(2,4)};

Step 2 includes the following 5 steps:

Step 2.1: A navel orange color image sample set is established through 50 navel orange images as shown in FIG2. It can be seen from FIG2 that some navel orange images do not contain immature areas, and some navel orange images also have various types of surface defects. Then, the color card number-related areas of the background, maturity, immaturity, and surface defects in each navel orange image are detected and merged. The color card number range related to maturity can be set to 122 to 124 and 135 to 138, the color card number range related to immaturity can be set to 372 to 377 and 379 to 382, the color card number range related to surface defects can be set to 443 to 457 and 539 to 544, and the color card number range related to background can be set to 577 to 580; thus, the historical thresholds of the three color card numbers can be set and They are 138, 382 and 544 respectively, and the smallest color card number in the navel orange color image can be set. and the largest color card number 122 and 580 respectively; the three color cards are numbered as historical thresholds and As the initial values of the color card number thresholds Γ(x ₁ ,y ₁ ,z ₁ ), Γ(x ₂ ,y ₂ ,z ₂ ) and Γ(x ₃ ,y ₃ ,z ₃ ) in the processing of the navel orange color image shown in the current FIG. 3( a ), the color card number threshold change is set to Δ=1, the loop process count t is 0, the threshold combination (Γ(x ₁ ,y ₁ ,z ₁ ),Γ(x ₂ ,y ₂ ,z ₂ ),Γ(x ₃ ,y ₃ ,z ₃ )) is included in the threshold permutation combination set ν(t=0), the threshold combination in ν(t=0) is calculated according to the specific form of the maximum inter-class variance of the three thresholds of the color card number described in step 1, and the calculation result is included in the maximum inter-class variance value set υ(t=0) of the three thresholds of the color card number;

Step 2.2: Add 1 to t instead of the original t, set the intermediate variable θ to Δ·t or 0, and classify the threshold combinations (Γ(x ₁ ,y ₁ ,z ₁ )±θ,Γ(x ₂ ,y ₂ ,z ₂ )±θ,Γ(x ₃ ,y ₃ ,z ₃ )±θ) into the threshold permutation combination set ν(t);

Step 2.3: Calculate the threshold combination in ν(t) for the navel orange color image shown in FIG. 3(a) according to the specific form of the maximum inter-class variance of the three thresholds of the color card number described in step 1, and include the calculation result for the navel orange color image shown in FIG. 3(a) into the maximum inter-class variance value set υ(t) of the three thresholds of the color card number;

Step 2.4: Compare υ(t) with υ(t-1) to see if there is a larger value. If υ(t) is larger than υ(t-1), go to step 2.2, otherwise go to step 2.5;

Step 2.5: The largest value in υ(t) corresponds to and The threshold combination is set to three optimized thresholds for dividing four color card number sets of Ω ₁ , Ω ₂ , Ω ₃ and Ω _4. Since FIG. 3(a) does not include an immature area, but there are two surface defect areas with greatly different features, it can be reduced to three color card number sets in step 2. It is also possible to divide the surface defects into two color card number sets by splitting the color card number range related to the surface defects, thereby improving the recognition of the surface defect extraction. In the embodiment of the present invention, the surface defects are divided into two color card number sets as an example, so the three optimized thresholds are and Compared with the historical thresholds of the three color card numbers and Adjustable to 137, 457 and 541;

Step 3: For the pixel point (m, n) in the navel orange color image in FIG3(a), the corresponding color card number set number f(m, n) can be determined by the element combination (l, a, b) corresponding to the pixel point (m, n), and the specific form is as follows:

The color card number set corresponding to all pixels in the output image is numbered, and the division of the four color card number sets Ω ₁ , Ω ₂ , Ω ₃ and Ω ₄ is completed. It can be seen that Ω ₁ and Ω ₄ are the color card number sets corresponding to the mature and background respectively, thereby extracting the color card number sets corresponding to the two surface defects Ω ₂ and Ω ₃ .

For the navel orange image shown in FIG3(a), the color card number set numbers corresponding to all pixels in the image can be output in the specific form of the color card number set numbers. FIG3(b) shows that the navel orange image is divided into four color card number sets. Since the navel orange image does not include an immature area, and the characteristics of the two surface defect areas are quite different, they can be split into two color card number sets for obvious distinction to further improve the recognition capability, so that the green and khaki marked areas are both surface defect areas, so as to realize surface defect extraction for fruit and vegetable post-harvest processing equipment.

Any part not specified in this embodiment can be implemented using existing technologies.

For those skilled in the art, according to the teachings of the present invention, without departing from the principles and spirit of the present invention, changes, modifications, substitutions and variations made to the implementation methods are still within the protection scope of the present invention.

Claims (3)

1. The surface defect extraction method for the fruit and vegetable postharvest treatment equipment is characterized by comprising the following steps of:

Step 1: providing a fruit and vegetable color image to be subjected to surface defect extraction for fruit and vegetable post-picking processing equipment, carrying out brightness normalization on each pixel point in the fruit and vegetable color image according to a certain color model, then outputting a certain color card number corresponding to three elements of the color model, setting the obtained color card number into four color card number sets which are classified into background, mature, immature and surface defect, providing probability and corresponding element average value generated by the four color card number sets, providing probability and corresponding element average value jointly generated by different color card number sets, providing adjacent color card number average element difference mutual information quantity when color card number difference is within a limiting condition through the color card number set and the next color card number set, providing maximum probability color card number average element difference mutual information quantity when the probability of each pixel point in the image is maximum through the color card number of the color card number set and the other color card number set, and determining three maximum color card number cross-class threshold values weighted by the comprehensive adjacent color card number average element mutual information quantity and the maximum probability color card number average element difference mutual information quantity of the four color card number sets;

Step 2: on the basis of taking the historical threshold values of three color card numbers as initial values, continuously increasing and decreasing the product of the color card number threshold value variation quantity and the cycle process count by the threshold value combination to be classified into a threshold value arrangement combination set, classifying the calculation result into a color card number three-threshold value maximum inter-class variance value set, and if the color card number three-threshold value maximum inter-class variance value set is not added with a larger value, dividing the threshold value combination corresponding to the largest value in the color card number three-threshold value maximum inter-class variance value set into three optimized threshold values of four color card number sets;

Step 3: for pixel points in the image, comparing the color card number corresponding to the element combination of the pixel point with three optimized thresholds for dividing four color card number sets in the step 2 to determine the corresponding color card number set number, and then outputting the color card number set numbers corresponding to all the pixel points in the image to complete the division of the four color card number sets so as to extract the surface defects;

in the step 1, each pixel point in the fruit and vegetable color image is normalized according to a certain color model, and then a certain color card number corresponding to three elements of the color model is output, and the color card number obtained by setting can be classified into a set of 4 color card numbers of background, mature, immature and surface defects, and the specific method is as follows:

the color card number corresponding to the element combination (l, a, b) after brightness normalization according to a certain color model in the fruit and vegetable color image can be obtained from To the point ofWherein l, a and b are respectively the values of three elements given by pixel points in the fruit and vegetable color image according to the color model,Respectively takes the values of three elements corresponding to the color with the minimum color card number in the fruit and vegetable color image,The color card number is respectively the value of three elements corresponding to the color with the largest color card number in the fruit and vegetable color image, and is divided into 4 color card number sets which are respectively characterized as background, mature, immature and surface defect, wherein the specific forms are as follows:

Ω₂＝{Γ(x₁,y₁,z₁)+1,...,Γ(x₂,y₂,z₂)}

Ω₃＝{Γ(x₂,y₂,z₂)+1,...,Γ(x₃,y₃,z₃)}

wherein Γ is an input element combination and a function ,(x₁,y₁,z₁)、(x₂,y₂,z₂)、(x₃,y₃,z₃) for outputting a color chart number is an element combination corresponding to a color chart number of a color chart number set of which Ω ₁、Ω₂、Ω₃ and Ω ₄ are divided;

the specific method for giving the probability generated by the four color card number sets and the average value of the corresponding elements in the step 1 is as follows:

The probabilities generated by the omega ₁、Ω₂、Ω₃ and omega ₄ color card number sets are kappa ₁、κ₂、κ₃ and kappa ₄ respectively, and the specific forms are as follows:

Wherein, i epsilon [1,2,3,4], epsilon _(l,a,b) is the probability of the element combination (l, a, b) in the pixel point of the image;

The element average values of the corresponding omega ₁、Ω₂、Ω₃ and omega ₄ color card number sets are respectively AndThe element average value corresponding to omega _i isThe specific form is as follows:

Wherein, alpha, beta and gamma are the weight coefficients of elements L, A and B in the element average value respectively;

The specific method for providing the probability and the average value of the corresponding elements generated by the different color card number sets in the step1 is as follows:

The probability of co-occurrence of Ω ₁ and Ω ₂ is κ _1,2,Ω₁ and Ω ₃, the probability of co-occurrence of κ _1,3,Ω₁ and Ω ₄ is κ _1,4,Ω₂ and Ω ₃, the probability of co-occurrence of κ _2,3,Ω₂ and Ω ₄ is κ _2,4,Ω₃ and Ω ₄ is κ _3,4, and the probability of co-occurrence of Ω _i and Ω _j is κ _i,j, in the following specific form:

Wherein i, j e [1,2,3,4], i not equal to j and i < j;

The average value of the elements corresponding to omega _i and omega _j is ψ _i,j, wherein i, j epsilon [1,2,3,4], i not equal to j and i < j, and the specific forms are as follows:

in the step 1, the specific method for giving the mutual information quantity of the average element difference of the adjacent color card numbers when the color card number difference is within the limiting condition through a certain color card number set and the next color card number set is as follows:

For Ω _i and Ω _j that satisfy i, j e [1,2,3,4] and i+1=j, setting the limiting value of the adjacent color card number difference as k, calculating the limiting condition Γ (x, y, z) - Γ (x ', y', z ') for the average element difference mutual information amount ζ _i,j,k of adjacent color card numbers of Ω _i and Ω _j in the limiting condition Γ (x, y, z) and any color card number Γ (x', y ', z') in the adjacent Ω _j by any color card number Γ (x, y, z) in Ω _i, specifically the following form:

Wherein, Is an arbitrary symbol;

In the step 1, the specific method for giving the maximum probability color card number average element difference mutual information quantity through the color card number when the probability of each pixel point in the image is maximum in a certain color card number set and another color card number set is as follows:

For omega _i and omega _j satisfying i, j ε [1,2,3,4] and j > i+1, the maximum probability color chip number average element difference mutual information amount for omega _i and omega _j can be calculated from color chip number Γ (x _i,y_i,z_i) when the probability of pixel point in omega _i is maximum and color chip number Γ (x _j,y_j,z_j) when the probability of pixel point in omega _j is maximum The specific form is as follows:

Wherein,

In the step 1, the specific method for determining the color card number three-threshold maximum inter-class variance weighted by the comprehensive adjacent color card number average element difference mutual information quantity and the maximum probability color card number average element difference mutual information quantity of the four color card number sets is as follows:

Determining the average element difference mutual information quantity xi _i,j,k and the maximum probability color card number average element difference mutual information quantity for adjacent color card numbers of omega ₁、Ω₂、Ω₃ and omega ₄ The weighted color chart number three-threshold maximum inter-class variance is σ²[Γ(x₁,y₁,z₁),Γ(x₂,y₂,z₂),Γ(x₃,y₃,z₃)], in the specific form as follows:

Where Θ ₁＝{(1,2),(2,3),(3,4)},Θ₂ = { (1, 3), (1, 4), (2, 4) }.

2. The method for extracting surface defects of fruit and vegetable post-harvest processing equipment according to claim 1, wherein the specific steps of step 2 are as follows:

step 2.1: numbering three color cards to a past history threshold AndAs initial values of color card number thresholds Γ (x ₁,y₁,z₁)、Γ(x₂,y₂,z₂) and Γ (x ₃,y₃,z₃) in the current fruit and vegetable color image processing, setting the variation of the color card number threshold as delta, setting the cyclic process count t as 0, classifying (Γ(x₁,y₁,z₁),Γ(x₂,y₂,z₂),Γ(x₃,y₃,z₃)) threshold combinations into a threshold arrangement combination set ν (t=0), calculating threshold combinations in ν (t=0) according to the specific form of the color card number three-threshold maximum inter-class variance in step 1, and classifying calculation results into the color card number three-threshold maximum inter-class variance value set ν (t=0);

Step 2.2: after adding 1 to t, replacing the original t, setting the value of an intermediate variable theta as delta t or 0, and classifying (Γ(x₁,y₁,z₁)±θ,Γ(x₂,y₂,z₂)±θ,Γ(x₃,y₃,z₃)±θ) threshold combinations into a threshold arrangement combination set v (t);

Step 2.3: calculating a threshold combination in v (t) according to the specific form of the color card number three-threshold maximum inter-class variance in the step 1, and classifying the calculation result into a color card number three-threshold maximum inter-class variance value set v (t);

step 2.4: comparing the upsilon (t) with the upsilon (t-1) to see whether larger numerical values appear, if so, entering the step 2.2, otherwise, entering the step 2.5;

step 2.5: corresponding the largest value in the v (t) AndThe threshold combination is set to three optimized thresholds that divide the sets of Ω ₁、Ω₂、Ω₃ and Ω ₄ color chart numbers.

3. The method for extracting surface defects of fruit and vegetable post-harvest processing equipment according to claim 1, wherein the step 3 specifically comprises:

For a pixel (m, n) in an image, a corresponding color card number set number f (m, n) can be determined by a corresponding element combination (l, a, b) of the pixel (m, n), and the specific form is as follows:

And outputting color card number sets corresponding to all pixel points in the image, and finishing the division of the omega ₁、Ω₂、Ω₃ and omega ₄ color card number sets so as to extract the color card number set corresponding to the surface defect.