CN113536687A - Construction method of color and dye concentration prediction model of monochrome cotton fabric based on PSO-LSSVM - Google Patents

Abstract

The invention provides a method for constructing a color and dye concentration prediction model for monochrome cotton fabrics based on PSO-LSSVM, comprising the following steps: obtaining qualified dyeing data of monochrome fabrics, dividing them into training sets and test sets; setting initial LSSVM parameters And establish the LSSVM initial model to predict the training set; use the PSO algorithm to perform parameter optimization processing on the prediction results and assign them to the LSSVM initial model to obtain the PSO-LSSVM prediction basic model; The test set is used as a test sample, and the PSO-LSSVM prediction base is used. The model is tested and trained to obtain the final PSO-LSSVM prediction model; the final PSO-LSSVM prediction model with the dye concentration as and the color feature parameters L*, a* and b* as the input and output values are the color prediction model and the dye concentration prediction model. . The color prediction model and the dye concentration prediction model established by the invention maintain the characteristic parameter values within the ideal range, and have good specificity.

Description

Construction method of color and dye concentration prediction model of monochrome cotton fabric based on PSO-LSSVM

technical field

The invention belongs to the field of dyeing and color matching, and particularly relates to a method for constructing a PSO-LSSVM-based monochromatic cotton fabric color and dye concentration prediction model.

Background technique

The process of pad-drying and pad-steaming is a common dyeing process for cotton fabrics. In the actual preparation of samples for customer confirmation, the method of manual color matching is still mainly used. The color proofing is prepared with the experience of color matching personnel. The cycle is long, and when batch dyeing Many unpredictable disturbances and variations are often encountered, resulting in low reproducibility and reduced productivity. With the rapid development of computer technology, mathematics, physics and other disciplines, the theory, method and technology of color matching are constantly changing and developing, and the computer color matching system also emerges as the times require. At present, there are mainly three computer color matching methods, namely, color number file retrieval, full spectrum color matching and tristimulus value color matching. Among them, the full spectrum color matching and triple stimulus value color matching are suitable for computer numerical calculation.

Machine learning methods have been successfully applied in many leaders, and computer color matching technology is a new generation of color matching technology. It successfully combines computer and color matching technology in the printing and dyeing industry, greatly improving the efficiency of color matching. Among the existing color matching methods, tristimulus color matching is the most common method. At present, it is basically based on the least squares method of polynomial regression and BP neural network to establish a color matching model between dye concentration and color tristimulus values. Computer color matching.

In the machine learning method, the support vector machine (SVM) based on the VC dimension (Vapnik-ChervonenkisDimension) and the principle of structural risk minimization in statistical learning theory is also a very important method, which can better solve the problem of small samples, nonlinear , high dimensionality and local minima have been successfully applied in many fields such as pattern recognition, artificial intelligence, medicine, and economics. But SVM shows high performance when dealing with limited training samples, however, when the amount of training data is large, the computational process becomes very difficult, which makes it time-consuming and unsuitable for real-time applications. Least Squares Support Vector Machine (LSSVM) can overcome the insufficiency of standard support vector machine, and transform the solution of the quadratic programming problem of SVM into the solution of linear equations. Therefore, LSSVM can greatly reduce the computational complexity when solving complex problems with big data and many factors, so it is faster and more accurate than SVM. It has the advantages of strong chemical ability and short training time. Although LSSVM has been applied to textile dyeing, there are few researches on color matching model based on LSSVM.

SUMMARY OF THE INVENTION

An object of the present invention is to provide a PSO-LSSVM-based method for building a color and dye concentration prediction model for single-color cotton fabrics, and to provide at least the advantages that will be described later.

A new and accurate method is proposed for computer color matching of cotton fabrics dyed with reactive dyes. The present invention is a method for predicting dye concentration based on PSO-LSSVM and a single reactive dye database. The method is based on the PSO-LSSVM algorithm and a single reactive dye database. The color prediction model and the dye concentration prediction model established by the reactive dye database, the characteristic parameter values of the two types of models can be maintained within the ideal range, and the constructed model and the single reactive dye database have good dye specificity, and can be used in predicting unknown single-color dyed fabrics. When formulating, the prediction error of dye concentration is small, and the formula is accurate. A new method is proposed for computer color matching.

The technical scheme of the present invention is as follows:

A method for building a color and dye concentration prediction model for single-color cotton fabrics based on PSO-LSSVM, which includes the following steps:

Obtaining qualified dyeing data for single-color fabrics, including fabric dye concentration and color characteristic parameters L*, a* and b* values, and dividing the fabric qualified dyeing data into a training set and a test set;

Set initial LSSVM parameters and establish an initial LSSVM model, and utilize the initial LSSVM model to predict the training set;

赋值给LSSVM初始模型，得到PSO-LSSVM预测基础模型；The PSO algorithm is used to optimize the parameters of the above prediction results, and the global optimal value gbest _d is obtained, that is, the optimal parameter group

Assign it to the LSSVM initial model to obtain the PSO-LSSVM prediction basic model;

With described test set as test sample, utilize PSO-LSSVM prediction basic model team to carry out test training, obtain final PSO-LSSVM prediction model;

The final PSO-LSSVM prediction model with dye concentration as input and color feature parameters L*, a* and b* values as output is the color prediction model; color feature parameters L*, a* and b* values of dyed fabrics are input dyes The final PSO-LSSVM prediction model whose concentration is the output is the dye concentration prediction model.

Preferably, in the described PSO-LSSVM-based method for constructing a color and dye concentration prediction model for monochromatic cotton fabrics, obtaining qualified dyeing data for monochromatic fabrics includes:

Preset the dyeing conditions and dyeing concentration when dyeing cotton fabrics with reactive dyes, and carry out the pad-bake pad-steam dyeing test to obtain the dyed color samples;

Use a spectrophotometer to test the color characteristic parameters of dyeing samples, remove abnormal dyeing data, re-dye, and record the dye concentration and color characteristic parameters L*, a* and b* values under qualified dyeing conditions;

in,

The dyeing conditions are as follows: the concentration of sodium chloride is 150-200g/L, sodium carbonate is 15-25g/L, sodium hydroxide is 4-8g/L, and the liquid rate of the fabric after padding is 60-75%. Pre-bake for 100-200s, steam for 120-250s under saturated steam at 100-102℃;

The dye concentration is 0.01-20g/L and the total does not exceed 20g/L, and it is divided into 15-30 dyeing groups according to different concentrations;

The dyeing conditions of each dyeing sample are the same;

Removing abnormal dyeing data includes using a desktop spectrophotometer to measure the dyed color sample, and recording the spectral reflectance curve and K/S value curve of the blank fabric and each dye mass concentration gradient color sample. Re-proofing and correction of irregular curves such as dislocation of absorption peaks.

Preferably, in the described PSO-LSSVM-based method for constructing a color and dye concentration prediction model for monochrome cotton fabrics, the PSO algorithm includes:

Set specific parameters for the dyeing effect prediction model to be established: population size m, particle dimension n, maximum position X _max , maximum velocity V _max , inertia weight w, learning factors c ₁ and c ₂ , maximum number of iterations K _max and The fitness function fitness() of the optimization problem;

Set the initial information of the population, including position _xi and velocity _vi ;

Calculate the fitness value of each particle according to the objective function, and evaluate the fitness of each particle;

与其历史最佳位置的适应 度值fitness(pbest_id)对比，若当前位置适应度值较高，则用当前位置更新历史最佳位置

Select the optimal position of the current individual and the current fitness value of each particle

Compared with the fitness value fitness(pbest _id ) of its historical best position, if the fitness value of the current position is higher, the current position is used to update the historical best position

与群体最佳位置的适应度 值fitness(gbest_d)对比，若当前位置适应度值较高，则用当前位置更新群体最佳位置

判断是否满足终止条件，满足则搜索停止，输出结果；不满足则继续下一步；Select the optimal position of the current group, and the current fitness value of each particle

Compared with the fitness value fitness(gbest _d ) of the best position of the group, if the fitness value of the current position is higher, the best position of the group is updated with the current position

Determine whether the termination condition is satisfied, if it is satisfied, the search will stop and the result will be output; if it is not satisfied, continue to the next step;

Determine whether the iteration end condition is met, and set the iteration end condition according to the optimization problem. If the end condition is not met, return to the second step to recalculate;

If the output condition is satisfied, output the global best position gbest _d and stop searching.

Preferably, in the described PSO-LSSVM-based method for constructing a color and dye concentration prediction model for monochrome cotton fabrics, the test set is used as a test sample, and the PSO-LSSVM is used to predict the basic model team for test training to obtain the final PSO -LSSVM prediction model includes:

According to the model characteristic parameter values RMSE, E _p , MAE and R ² , the quality of the model is comprehensively judged;

If the established prediction model does not reach the predetermined value, return to the construction of the color prediction model and the dye concentration prediction model again;

Among them, the calculated value of RMSE should be less than 0.40, the calculated value of MAE should be less than 0.35, the calculated value of Ep should be greater than 90%, and the calculated value of R ² should be greater than 99%.

Preferably, in the described PSO- ^LSSVM -based method for constructing a color and dye concentration prediction model for monochrome cotton fabrics, MAE, RMSE, _Ep and R are defined as:

是预测值，

是样本平均值。where n is the number of samples, y is the experimental value,

is the predicted value,

is the sample mean.

The above algorithm principles and specific steps involving LSSVM are: given a set of training data sets {(xi, yi), i=1, 2, N}, where N is the number of training samples; ^xi∈Rd is the input matrix, d is the dimension of the matrix; yi ∈ R is the output matrix. Then, through a nonlinear mapping φ(x), the training dataset is mapped from the original space to the high-dimensional feature space. The decision function of LSSVM is as follows:

where w is the weight vector, and the constant b is the bias term. The optimization problem of LSSVM is as follows:

In the formula: γ is the regularization parameter, ξ is the error variable. Compared with the SVM algorithm, its constraints are different, and it becomes the following form:

Change the inequality constraints of the support vector machine to equality constraints, and construct the Lagrangian function L to solve:

where αi is the Lagrange multiplier. According to the Karush-Kuhn-Tucker (KKT) condition, the following equation constraints are obtained:

Next, eliminate w and ξ to get the following linear system of equations:

In the formula: I is the identity matrix; α is the support vector; others are as follows:

Use Ω to represent the inner product of the kernel matrix, and the kernel function technique is as follows:

Finally, the decision function of LSSVM, that is, the regression model, is obtained as follows:

In the formula: α and b are obtained by solving. In the actual modeling process, α and b are obtained by applying the LSSVM toolbox programming operation in the MATLAB platform, and the support vector α is a matrix with N rows and 1 column.

The present invention has the following beneficial effects:

Based on the color prediction model and dye concentration prediction model established by the PSO-LSSVM algorithm, the characteristic parameter values of the two models can be easily maintained within the ideal range, and the specificity is good;

The prediction error of dye concentration is small, and the prediction of double spell dyeing formula is accurate;

There are few test groups required to obtain qualified dyeing data, and the establishment of a database is simple and has great application prospects.

Other advantages, objects, and features of the present invention will appear in part from the description that follows, and in part will be appreciated by those skilled in the art from the study and practice of the invention.

Description of drawings

Fig. 1 is the flow chart of the method for building a monochrome cotton fabric color and dye concentration prediction model based on PSO-LSSVM provided by the invention;

Fig. 2 is the POS algorithm flow chart in one embodiment of the method for constructing the color and dye concentration prediction model of monochrome cotton fabric based on PSO-LSSVM provided by the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings, so that those skilled in the art can refer to the description and implement accordingly.

It should be understood that terms such as "having", "comprising" and "including" as used herein do not assign the presence or addition of one or more other elements or combinations thereof.

As shown in Figure 1, the present invention provides a kind of monochromatic cotton fabric color and dye concentration prediction model construction method based on PSO-LSSVM, comprises the following steps:

S1: Design the dyeing conditions and a series of dye concentrations when dyeing cotton fabrics with reactive dyes, and conduct a pad-bake pad-steam dyeing test;

S2: Test the color characteristic parameters CIE Lab of cotton fabrics of each group of dyeing conditions, remove abnormal data, record the dye concentration and color characteristic parameters L*, a* and b* values under qualified dyeing conditions, and establish a single reactive dye database;

S3: randomly select 4/5 data from the single reactive dye database established in the above S2 as a training group, and establish a color prediction model with the dye concentration as the input and the color characteristic parameters L*, a* and b* of the dyed fabric as the output respectively. , and a dye concentration prediction model that takes the color characteristic parameters L*, a* and b* of dyed fabrics as the input dye concentration as the output;

S4: Use the remaining data as the test group, and comprehensively judge the quality of the model according to the model characteristic parameter values RMSE, _Ep , ^MAE and R2. If the model has no specified quality requirements, return to S3;

S5: Measure the unknown single-color dyed cotton fabric with a spectrophotometer, and input the measured color characteristic parameters L*, a* and b* values into the formula prediction model to obtain the dye concentration;

S6: then input the dye concentration into the color prediction model to obtain the color characteristic parameters L*, a* and b* values, and calculate the color difference ΔE _ab * according to the experimental color sample and the predicted color characteristic parameters;

S7: Compare each single reactive dye database and the calculated color difference ΔE _ab * under the corresponding prediction model, and the dye concentration with the smallest color difference value is the best predicted dye concentration.

In the above-mentioned step S1, the dyeing conditions are as follows: the concentration of sodium chloride is 150-200g/L, the sodium carbonate is 15-25g/L, the sodium hydroxide is 4-8 g/L, and the liquid rate of the fabric after padding is 60-75%. Pre-bake under hot air at 110-125°C for 100-200s, and steam under saturated steam at 100-102°C for 120-250s; the dye concentration in the above step S1 is 0.01-20g/L, designed according to a certain concentration gradient, a total of 15-30 groups ;

In the above-mentioned single reactive dye database, each group of data is obtained under the same dyeing condition;

The method for removing abnormal data in the above step S2 is: use a desktop spectrophotometer to measure the color of the dyed color sample in the above step S1, and record the spectral reflectance curve and K/S value of the blank fabric and each dye mass concentration gradient color sample curve, and re-proofing and correcting irregular curves such as curve crossing and absorption peak dislocation.

采用径向基核函数 (kernel＝′RBF_kernel′)，并以均方误差作为适应度函数(fitness()＝MSE)，设定精度 ep＝10^-3，满足设定精度或最大迭代次数则输出参数组，建立PSO-LSSVM预测模型。For the color prediction model and the dye concentration prediction model established in the above step S3, the parameters are set as: the number of particles popsize=20, the maximum number of iterations maxgen=200, the learning factor c ₁ =c ₂ =1.4, the inertia weight w∈[0.8, 1.2], regularization parameter γ∈[0.1, 1000], kernel function width σ ² ∈ [0.1, 100], regularization parameter γ iteration speed v _γ ∈ [-500, 500], kernel function width σ ² iteration speed

The radial basis kernel function (kernel='RBF_kernel') is used, and the mean square error is used as the fitness function (fitness()=MSE), and the set precision is ep=10 ^-3 . If the set precision or the maximum number of iterations is satisfied, the output will be Parameter group, establish PSO-LSSVM prediction model.

In the above step S4, the calculated value of the model feature parameter RMSE is less than 0.40, the calculated value of MAE is less than 0.35, the calculated value of _Ep is greater than 90%, and the calculated value of R ² is greater than 99%. MAE, RMSE, _Ep and R ² are defined as:

是预测值，

是样本平均值。where n is the number of samples, y is the experimental value,

is the predicted value,

is the sample mean.

The calculation formula of ΔE _ab * in the above step S7 is as follows:

In the formula: ΔE _ab * is the total color difference between the experimental color sample and the predicted color; ΔL*, Δa* and Δb* are the CIELAB color space, and the calculation formula is as follows:

L ^* =116f(Y/ _Yn )-16

a ^* =500[f(X/ _Xn )-f(Y/ _Yn )]

b ^* =200[ _f (Y/ _Yn )-f(Z/Zn)]

Wherein, X/X _n , Y/Y _n , and Z/Zn need to be greater than ( _6/29 ) ³ or less than or equal to (6/29) ³ at the same time.

In the above formula, X, Y, and Z are the _tristimulus values of the color sample; _Xn , _Yn , and Zn are the tristimulus values of the CIE standard illuminator irradiated on the surface of the completely diffuse reflector and then reflected to the observer's eyes.

Based on PSO-LSSVM, a color prediction model and a dye concentration prediction model are established. The modeling process is as follows:

(1) Judging whether the dyeing effect parameters and dyeing influencing factor parameters need to be normalized, forming a sample matrix, and dividing the training set and the test set;

(2) Set the range of the RBF kernel function parameter group (γ, σ ² ), usually γ∈[0.1, 1000], σ2∈[0.1, 100]. Set the PSO related parameters, and use the PSO algorithm to initialize the population;

(3) Determine the fitness function, calculate and compare the particle fitness value with the training set, select the individual optimal value pbest _id and the global optimal gbest _d , and update the position and speed of each particle;

(4) Iterative optimization to meet the end condition (minimum fitness value or maximum number of iterations);

赋值给LSSVM，导入训练集， 调用trainlssvm函数对所得参数进行训练，得到训练好的PSO-LSSVM模型；(5) Output the global optimal value gbest _d , that is, the optimal parameter group

Assign the value to LSSVM, import the training set, call the trainlssvm function to train the obtained parameters, and obtain the trained PSO-LSSVM model;

(6) Call the simlssvm function to simulate the trained model, substitute the input matrix of the test set into the prediction model, obtain the predicted value of the output matrix of the test set through the simulation operation, and use the evaluation index to compare with the experimental value, if the simulation effect reaches the expectation Then go to the next step, otherwise continue to use the PSO algorithm to optimize the parameter group.

As shown in Figure 2, the above PSO algorithm specifically includes the following steps:

(1) First, specific parameters need to be set for the prediction model to be established: group size m, particle dimension n, maximum position X _max , maximum velocity V _max , inertia weight w, learning factors c ₁ and c ₂ , maximum number of iterations K _max and the fitness function fitness() of the optimization problem, and then set the initial information of the population, including the position x _i and the speed v _i ;

(2) Calculate the fitness value of each particle according to the objective function, and evaluate the fitness of each particle;

与其历史最佳位置 的适应度值fitness(pbest_id)对比，若当前位置适应度值较高，则用当前位置更新历史最佳位置

(3) Select the optimal position of the current individual, and the current fitness value of each particle

Compared with the fitness value fitness(pbest _id ) of its historical best position, if the fitness value of the current position is higher, the historical best position is updated with the current position

与群体最佳位置的 适应度值fitness(gbest_d)对比，若当前位置适应度值较高，则用当前位置更新群体最佳位置

判断是否满足终止条件，满足则搜索停止，输出结果；不满足则继续下一步；(4) Select the optimal position of the current group, and the current fitness value of each particle

Compared with the fitness value fitness(gbest _d ) of the best position of the group, if the fitness value of the current position is higher, the best position of the group is updated with the current position

Determine whether the termination condition is satisfied, if it is satisfied, the search will stop and the result will be output; if it is not satisfied, continue to the next step;

(5) Judging whether the iteration end condition is met, and setting the iteration end condition according to the optimization problem, if the end condition is not met, return to the second step for recalculation;

(6) If the output condition is satisfied, output the global best position gbest _d and stop searching.

The above algorithm principles and specific steps involving LSSVM are: given a set of training data sets {(xi, yi), i=1, 2, N}, where N is the number of training samples; ^xi∈Rd is the input matrix, d is the dimension of the matrix; yi ∈ R is the output matrix. Then, through a nonlinear mapping φ(x), the training dataset is mapped from the original space to the high-dimensional feature space. The decision function of LSSVM is as follows:

where w is the weight vector, and the constant b is the bias term. The optimization problem of LSSVM is as follows:

In the formula: γ is the regularization parameter, ξ is the error variable. Compared with the SVM algorithm, its constraints are different, and it becomes the following form:

Change the inequality constraints of the support vector machine to equality constraints, and construct the Lagrangian function L to solve:

where αi is the Lagrange multiplier. According to the Karush-Kuhn-Tucker (KKT) condition, the following equation constraints are obtained:

Then, by eliminating w and ξ from equation (3.5), the following linear equations are obtained:

In the formula: I is the identity matrix; α is the support vector; others are as follows:

Use Ω to represent the inner product of the kernel matrix, and the kernel function technique is as follows:

Finally, the decision function of LSSVM, that is, the regression model, is obtained as follows:

In the formula: α and b are obtained by solving the formula (3.6). In the actual modeling process, α and b are obtained by applying the LSSVM toolbox programming operation in the MATLAB platform, and the support vector α is a matrix with N rows and 1 column.

Dyeing with Lihua Red, using pad-bake-pad-steam dyeing process: pad dyeing solution (two dips and two pads, with liquid rate 60%) → pre-baking 120s (120 ℃ hot air) → padding fixing solution (two dips and two rollings, with liquid rate 70%, sodium chloride 180g/L, sodium carbonate 25g/L, sodium hydroxide 6g/L) → steaming for 150s (100～102℃, saturated steam) → cold water washing → hot Water washing→soap washing (standard soap flakes 3g/L, bath ratio 1:50, 90℃ for 10min)→hot water washing→cold water washing→drying.

A total of 25 dye mass concentration gradients (0.01, 0.05, 0.10, 0.50, 1.00, 1.25, 1.5, 1.75, 2.00, 2.50, 3.00, 3.50, 4.00, 4.50, 5.00, 5.50, 6.00, 7.00, 8.00, 10.00, 12.00, 14.00, 17.00, 19.00, 20.00), the concentration unit is g/L.

Use the desktop spectrophotometer X-Rite Color I5 to measure the dyed color samples, record the spectral reflectance curve and K/S value curve of the blank fabric and each dye mass concentration gradient color sample. Irregular curves such as peak dislocation shall be re-proofed and corrected. Store all the information that conforms to the color sample, and establish a dedicated monochrome database for the establishment of color prediction models and color matching models.

Use the desktop spectrophotometer X-Rite Color I5 to measure the L* value, a* value and b* value of the color sample under the measuring aperture Φ6mm, D65 standard light source, and 10° viewing angle. Fold a color swatch into two layers, select a measurement point on each surface, a total of four measurement points, and use the average value of the data of the four measurement points to represent the color of the color swatch.

Based on PSO-LSSVM, a color prediction model and a dye concentration prediction model are established. The modeling process is as follows:

(1) Judging whether the dyeing effect parameters and dyeing influencing factor parameters need to be normalized, forming a sample matrix, and dividing the training set and the test set;

(2) Set the range of the RBF kernel function parameter group (γ, σ ² ), usually γ∈[0.1, 1000], σ ² ∈ [0.1, 100]. Set the PSO related parameters, and use the PSO algorithm to initialize the population;

(3) Determine the fitness function, calculate and compare the particle fitness value with the training set, select the individual optimal value pbest _id and the global optimal gbest _d , and update the position and speed of each particle;

(4) Iterative optimization to meet the end condition (minimum fitness value or maximum number of iterations);

赋值给LSSVM，导入训练集， 调用trainlssvm函数对所得参数进行训练，得到训练好的PSO-LSSVM染色效果预测模型；(5) Output the global optimal value gbest _d , that is, the optimal parameter group

Assign the value to LSSVM, import the training set, call the trainlssvm function to train the obtained parameters, and obtain the trained PSO-LSSVM staining effect prediction model;

(6) Call the simlssvm function to simulate the trained model, substitute the input matrix of the test set into the prediction model, obtain the predicted value of the output matrix of the test set through the simulation operation, and use the evaluation index to compare with the experimental value, if the simulation effect reaches the expectation Then go to the next step, otherwise continue to use the PSO algorithm to optimize the parameter group.

In the process of establishing the above-mentioned color prediction model and color prediction model, the following instructions need to be made:

(1) Input variable and output variable

For the color prediction model, the input variable is the dye mass concentration and the output variable is the L* value, a* value and b* value in the CIELAB uniform color space, since there are three output variables, there are three different PSO-LSSVMs for one dye model, that is, there are three sets of parameter groups.

The three parameter groups share a training set and a test set;

For the color prediction model, the input variables are the L* value, a* value and b* value in the CIELAB uniform color space, and the output variable is the dye mass concentration;

(2) Parameter setting

采用径向基核函数(kernel＝′RBF_kernel′)，并 以均方误差作为适应度函数(fitness()＝MSE)，设定精度ep＝10^-3，满足设定精度或最大 迭代次数则输出参数组，建立PSO-LSSVM预测模型。After the PSO parameters have been verified by multiple simulations, the specific settings of the model are: particle number popsize=20, maximum iteration number maxgen=200, learning factor c ₁ =c ₂ =1.4, inertia weight w∈[0.8, 1.2], regularization parameter γ∈[0.1, 1000], kernel function width σ ² ∈ [0.1, 100], regularization parameter γ iteration speed v _γ ∈ [-500, 500], kernel function width σ ² iteration speed

The radial basis kernel function (kernel='RBF_kernel') is used, and the mean square error is used as the fitness function (fitness()=MSE), and the set precision is ep=10 ^-3 . If the set precision or the maximum number of iterations is satisfied, the output will be Parameter group, establish PSO-LSSVM prediction model.

(3) Sample division

A single-color database has a total of 25 color samples, of which 20 groups are selected as training sets, and the remaining 5 groups are used as test sets. The color prediction model and the color matching model use the same training set and test set, but the input matrix and output matrix are different. Select 20 sets of data as PSO-LSSVM training data, and the remaining 5 sets of data as PSO-LSSVM test data; based on MATLAB R2016b software platform, combined with LSSVM toolbox (version 1.8), adjust the code, import the dyeing experimental data for dyeing model Establish. The 20 sets of data are paired with the PSO-LSSVM training data color prediction model, which contains 3 sub-models, which are PSO-LSSVM-L, PSO-LSSVM-a and PSO-LSSVM-b, respectively. Three sets of model parameters (γ, σ2 ); The PSO-LSSVM model is tested with the remaining 5 sets of data, and the 4 evaluation indexes MAE, RMSE, E _p and R ² of the color prediction model are obtained.

In the construction of the dye color matching prediction model, the training set and test set are the same as the color prediction model, but the input matrix and output matrix are different, and the L* value, a* value and b* value of the color parameter CIELAB of the dyed sample are used as input. Matrix with dye concentrations as output matrix. The same 20 sets of data are selected as the PSO-LSSVM training data, and the remaining 5 sets of data are used as the PSO-LSSVM test data. The modeling process is the same as the color prediction model.

Validation of the experimental effect of the color prediction model:

In order to further prove the reliability of the color prediction model, the color difference value ΔE _ab * was calculated to further test the accuracy of the dyeing prediction model. The predicted values of the L*, a*, and b* values of the color parameters of each set of test sets are called a set of predicted samples. Table 1 shows the color difference values and colors between the five groups of test samples and predicted samples of the color prediction model.

Table 1 Color parameters of Lihua real red test samples and predicted samples

For textile fabrics, when the color difference is 0≤ΔE _ab *≤0.5, the color difference is hardly felt, and when the color difference is 0.5<ΔE _ab *≤1.5, the color difference is very small. The color difference ΔE _ab * of the 5 sets of test data is in the range of [0.33, 0.94], and the color difference is all less than 1. It can be seen from Table 1 that in the group with the largest color difference, the biggest factor causing the error is the redness a*, the error is only 0.85, and the color difference between the experimental value and the predicted value can hardly be felt visually. The color difference results show that the prediction results of the color prediction model are very ideal, which are basically the same as the color parameters measured by the colorimeter.

Validation of experimental results of dye color matching prediction model:

Verification of the same dye: The predicted mass concentration of Lihua Red is dyed with the same dyeing process, and the same color parameter test method is used to obtain the L* value, a* value and b* value of the color parameter CIELAB of the dyed color sample, as shown in Table 2 shown.

Table 2 Color parameters of Lihua real red test samples and predicted samples

It can be seen from Table 2 that the color parameters of the Lihua real red mass concentration dyed color samples predicted based on the dye color matching prediction model are very close to the color parameters of the test samples, and the color difference ΔE _ab * of the five sets of test data is in [0.46, 1.22] Within the range, all are less than 1.5, and the color difference feels very small.

Verification of different dyes: Use dyes with similar hues for testing (here, Remazol Red RGB), obtain test color samples after dyeing, measure color parameters, and substitute them into the predicted concentration obtained by the dye color matching prediction model. The pad-baking pad-steaming process was used for dyeing, the color parameters were measured, and the color difference was compared with the primary color sample to test the specificity of the model, as shown in Table 3.

Table 3 PSO-LSSVM ₇ specificity verification

It can be seen from Table 3 that according to the prediction results of the dye color matching prediction model, the color parameters of the color samples obtained by dyeing and the color parameters of the primary color samples are very different, and the color difference ΔE _ab * is greater than 6.0. The abbreviation of ) The color difference standard shows that there is a large color difference between the color samples, and the color matching result is unacceptable, not to mention that the color matching color difference meets the requirement of less than 1.5. It shows that the dye color matching prediction model has high specificity and is not suitable for other dye color matching. Even if the same type of dyes with similar hues are used, a large error will occur.

As can be seen from the above: the present invention is a color prediction model and a dye concentration prediction model established based on the PSO-LSSVM algorithm and a single reactive dye database, the model prediction quality is high, the stability is good, and the generalization ability is strong. The constructed model and the single reactive dye database The specificity of dyes is good. When predicting the formula of unknown single-color dyed fabrics, the prediction error of dye concentration is small, and the formula is accurate. It is a new, efficient and accurate method for computer color matching.

Although the embodiment of the present invention has been disclosed as above, it is not limited to the application listed in the description and the embodiment, and it can be applied to various fields suitable for the present invention. For those skilled in the art, it can be easily Therefore, the invention is not limited to the specific details without departing from the general concept defined by the appended claims and the scope of equivalents.

Claims (5)

1. The PSO-LSSVM-based method for constructing the monochrome cotton fabric color and dye concentration prediction model is characterized by comprising the following steps of:

obtaining qualified dyeing data of a single-color fabric, wherein the qualified dyeing data comprises fabric dye concentration and color characteristic parameters L, a and b, and the qualified dyeing data of the fabric is divided into a training set and a testing set;

setting initial LSSVM parameters, establishing an LSSVM initial model, and predicting the training set by using the LSSVM initial model;

performing parameter optimization processing on the prediction result by using a PSO algorithm to obtain a global optimal value gbest_dI.e. optimizing parameter sets

Assigning to an LSSVM initial model to obtain a PSO-LSSVM prediction base model;

taking the test set as a test sample, and performing test training by using a PSO-LSSVM prediction basis model team to obtain a final PSO-LSSVM prediction model;

taking the dye concentration as input and the color characteristic parameters L, a and b as output to obtain a final PSO-LSSVM prediction model as a color prediction model; and taking the values of the color characteristic parameters L, a and b of the dyed fabric as input dye concentration as output, and taking the final PSO-LSSVM prediction model as a dye concentration prediction model.

2. The PSO-LSSVM based monochromatic cotton fabric color and dye concentration prediction model construction method according to claim 1, wherein obtaining monochromatic fabric qualified dyeing data comprises:

presetting dyeing conditions and dyeing concentration when the reactive dye dyes cotton fabrics, and carrying out a rolling, drying, rolling and steaming dyeing test to prepare a dyed sample;

testing the color characteristic parameters of the dyed sample by using a spectrocolorimeter, removing abnormal dyeing data, re-dyeing, and recording the dye concentration and the values of the color characteristic parameters L, a and b under the qualified dyeing condition;

wherein,

the dyeing conditions are that the concentration of sodium chloride is 150-200g/L, the concentration of sodium carbonate is 15-25g/L, the concentration of sodium hydroxide is 4-8g/L, the liquid carrying rate of the padded fabric is 60-75%, the fabric is pre-dried for 200s at the temperature of 110-102 ℃ and steamed for 250s at the temperature of 100-102 ℃ and saturated steam;

the concentration of the dye is 0.01-20g/L and the total concentration is not more than 20g/L, and the dye is divided into 15-30 dyeing groups according to different concentrations;

the dyeing conditions of all dyeing samples are the same;

and the abnormal dyeing data elimination comprises the steps of carrying out color measurement on the dyed color sample by using a table type light splitting color measuring instrument, recording the spectral reflectance curve and the K/S value curve of the blank fabric and each dye mass concentration gradient color sample, and carrying out re-sampling correction on irregular curves such as curve intersection, absorption peak dislocation and the like.

3. The PSO-LSSVM-based method for constructing a monochrome cotton fabric color and dye concentration prediction model according to claim 2, wherein the PSO algorithm comprises:

specific parameters are set according to a dyeing effect prediction model to be established: population size m, dimension n of particle, maximum position X_maxMaximum velocity V_maxInertia weight ω, learning factor c₁And c₂Maximum number of iterations K_maxAnd a fitness function of the optimization problem, fitness ();

setting initial information of population, including position x_iAnd velocity v_i；

Calculating the fitness value of each particle according to the objective function, and evaluating the fitness of each particle;

selecting the current individual optimal position, and for each particle, selecting the current fitness value

Fitness value of fitness (pbest) with its historical best position_id) Comparing, if the fitness value of the current position is higher, updating the historical best position by the current position

Selecting the optimal position of the current group, and for each particle, determining the current fitness value

Fitness value fitness (gbest) with the best position of the group_d) Comparing, if the fitness value of the current position is higher, updating the best position of the group by using the current position

Judging whether a termination condition is met, if so, stopping searching and outputting a result; if not, continuing the next step; judging whether an iteration end condition is met, setting the iteration end condition according to the optimization problem, and returning to the second step for recalculation if the iteration end condition is not met;

if the output condition is met, outputting the global maximumGood position gbest_dAnd stops the search.

4. The PSO-LSSVM based monochromatic cotton fabric color and dye concentration prediction model construction method as claimed in claim 3, wherein the step of taking the test set as a test sample and performing test training by utilizing a PSO-LSSVM prediction basis model team to obtain a final PSO-LSSVM prediction model comprises the following steps:

according to model characteristic parameter values RMSE, E_pMAE and R²Comprehensively judging the quality of the model;

if the established prediction model does not reach the preset value, returning to build the color prediction model and the dye concentration prediction model again;

wherein the calculated RMSE is less than 0.40, the calculated MAE is less than 0.35, and E_pCalculated value of greater than 90%, R²Calculated values are greater than 99%.

5. The PSO-LSSVM-based method for constructing a monochrome cotton fabric color and dye concentration prediction model according to claim 4, wherein MAE, RMSE, E_pAnd R²Is defined as:

where n is the number of samples, y is the experimental value,

is a predicted value of the number of the frames,

is the sample average.

Images