A color matching recommendation method and system

Abstract

The application provides a color matching recommendation method and system, and belongs to the field of textile printing and dyeing. The method comprises the following steps: establishing a color matching formula history database; obtaining a color matching data set based on the color matching formula history database and a decision tree classification model; training a pre-constructed LightGBM initial classification model based on the color matching data set to obtain a LightGBM classification model; obtaining sample color parameters, sample material parameters and sample process parameters according to a sample to be matched; obtaining a target formula database subset from the color matching formula history database based on the sample material parameters and the sample process parameters; inputting the sample color parameters into the LightGBM classification model to enable the LightGBM classification model to output a sample color system corresponding to the sample to be matched; obtaining a target color system subset from the target formula database subset based on the sample color system; and obtaining a color matching scheme meeting a preset requirement according to the sample color parameters and the target color system subset. The application can make the color matching process intelligent and improve the accuracy and recommendation efficiency of color matching formula recommendation.

Description

Technical Field

[0001] This invention relates to the textile printing and dyeing industry, and in particular to a color matching recommendation method and system. Background Technology

[0002] Color matching is a crucial production step in the textile printing and dyeing industry. Current color matching methods largely rely on manual experience, resulting in low efficiency. To date, the main problems with existing mainstream technologies are as follows: If manual experience is used for formula management, there is a risk of inconsistent formula management, as it relies primarily on handwritten or paper records, making retrieval and preservation difficult. Furthermore, the over-reliance on manual experience leads to low efficiency among novice technicians.

[0003] While existing technologies include intelligent color matching systems, the lack of research on establishing and matching historical formula databases can lead to recommendations that are not targeted, inefficient, and unclear in determining the matching results. For example, if there are no formulas with the same color difference in the historical data, the system cannot output a formula result. Alternatively, the formula concentration may need to be revised based on the actual color matching process, which is also inefficient and heavily reliant on human experience. Summary of the Invention

[0004] The present invention aims to provide a color matching recommendation method and system to solve the above-mentioned technical problems, thereby improving the accuracy and efficiency of color matching recommendation.

[0005] To address the aforementioned technical problems, this invention provides a color matching recommendation method, comprising the following steps: Establish a historical database of color matching formulas; Based on the historical database of color matching recipes and the preset decision tree classification model, several color matching datasets are obtained; The pre-constructed LightGBM initial classification model is trained based on several of the aforementioned color datasets to obtain the LightGBM classification model; Obtain the sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched; Based on the sample material parameters and sample process parameters, a subset of the target formula database is obtained from the color formula history database; Input the sample color value parameter into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched; Based on the sample color system, obtain a subset of the target color system from the subset of the target formula database; Based on the sample color value parameters and the target color system subset, obtain a color scheme that meets the preset requirements.

[0006] In the above scheme, a historical database of color matching formulas is first established by collecting existing color matching data and then filtered using a pre-defined decision tree classification model to obtain a color matching dataset. Based on this, a pre-built LightGBM initial classification model is trained to obtain a LightGBM classification model, and the color matching dataset is further divided into color system subsets. For a sample to be matched, its color value parameters, material parameters, and process parameters are first obtained. Based on the material and process parameters, a target formula database subset can be filtered from the historical color matching formula database. Then, the color parameters are input into the LightGBM classification model to determine the color system, and then the target color system subset is obtained from the target formula database subset. Finally, a color matching scheme that meets the preset requirements is obtained based on the color parameters and the target color system subset.

[0007] The above solution digitizes the dyeing formula data and uses decision tree classification model and LightGBM classification model to efficiently group and filter the formula data, thereby making the color matching process intelligent and automated, reducing reliance on human experience, and improving the accuracy and efficiency of color matching formula recommendations.

[0008] Furthermore, the step of obtaining several color matching datasets based on the historical color matching formula database and the preset decision tree classification model includes: extracting material features and process features of different color matching formulas from the historical color matching formula database to obtain a material feature set and a process feature set; inputting the material feature set and the process feature set into the decision tree classification model so that the decision tree classification model splits the historical color matching formula database into several color matching datasets based on the material feature set and the process feature set; The color matching dataset is a dataset composed of historical color matching formulas with the same material characteristics and the same process characteristics.

[0009] In the above scheme, to obtain a color matching dataset composed of historical color matching formulas with the same material and process characteristics, a decision tree classification model is used. The model filters the historical color matching formula database according to requirements such as fabric material and dyeing process parameters. A decision tree is a recursive tree structure that classifies or regresses data through a series of rules. The decision tree classification model divides the dataset into multiple subsets by selecting the optimal split point, maximizing (or decreasing) the purity (or impurity) of each subset. By filtering using a pre-set decision tree classification model, the historical color matching formula database can be split into several color matching datasets based on the material feature set and the process feature set for subsequent construction of the lightGBM model.

[0010] Further, the step of training the pre-constructed LightGBM initial classification model based on the several color matching datasets to obtain the LightGBM classification model includes: preprocessing the data based on the historical color matching recipe database and dividing the dataset into a training set and a test set; training the pre-constructed LightGBM initial classification model based on the training set and specifying parameters to obtain a training model; evaluating the performance and tuning the parameters of the training model based on a preset cross-validation method to obtain a validation model; and evaluating the validation model based on the test set to determine that the validation model meets the preset training requirements, thus obtaining the LightGBM classification model.

[0011] Furthermore, if the verification model does not meet the preset training requirements, the verification model is retrained based on the training set and specified parameters to obtain a new training model and a new verification model, until the new verification model meets the preset training requirements.

[0012] The above scheme first requires data preprocessing of the historical color formula database. This includes dividing the colors in the database into seven color systems (red, orange, yellow, green, cyan, blue, and purple) based on the CIELAB color model and OHTA color space. A pre-built LightGBM initial classification model is then established, and the dataset is divided into training and test sets. Next, the pre-built LightGBM initial classification model is trained using the training set and specified parameters to obtain a training model. During training, the performance of the training model is evaluated and parameters are fine-tuned using a pre-defined cross-validation method to obtain a validation model. This validation model is then evaluated using the test set to ensure it meets the pre-defined training requirements, thus obtaining the LightGBM classification model.

[0013] It should be noted that if the verification model does not meet the preset index requirements, the verification model should be retrained based on the training set and with specified parameters to obtain a new training model and a new verification model, until the new verification model meets the preset training requirements.

[0014] Further, the step of obtaining a color scheme that meets the preset requirements based on the sample color value parameters and the target color system subset includes: calculating the Euclidean distance between the sample color value parameters and each color value in the target color system subset to obtain a distance value sequence; obtaining a target distance value that meets the preset requirements from the distance value sequence, and using the target color system subset corresponding to the target distance value as the color scheme.

[0015] Further, the step of calculating the Euclidean distance between the sample color value parameter and the color value of each color system in the target color system subset to obtain a distance value sequence is specifically as follows: For any color system in the target color system subset, the Euclidean distance between its color system value and the sample color value parameter is calculated as follows: in ρ For Euclidean distance, R x , G x , B x The sample color value parameter, R n , G n , B n For any color system color value parameter in the target color system subset; The distance value sequence is a data sequence consisting of the distance values ​​between each color system color value and the sample color value parameter.

[0016] The Euclidean distance obtained here is used to characterize the color difference between the sample and each color system in the target color system subset. It should be noted that by sorting the sequence of distance values ​​obtained by traversing the sample color values ​​and the target color system subset from smallest to largest, the dyeing formulas corresponding to the first few distance values ​​can be selected as the proposed recommended formulas for output.

[0017] This invention also provides a color matching recommendation system, comprising: a database construction module for establishing a historical database of color matching recipes; a color matching classification module for acquiring several color matching datasets based on the historical database of color matching recipes and a preset decision tree classification model; a classification model training module for training a pre-constructed LightGBM initial classification model based on the several color matching datasets to obtain a LightGBM classification model; a sample parameter acquisition module for acquiring sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched; a target recipe screening module for acquiring a subset of target recipe databases from the historical database of color matching recipes based on the sample material parameters and sample process parameters; a sample color system acquisition module for inputting the sample color value parameters into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched; a target color system acquisition module for acquiring a subset of target color systems from the subset of target recipe databases based on the sample color systems; and a color matching recommendation module for acquiring color matching schemes that meet preset requirements based on the sample color value parameters and the subset of target color systems.

[0018] The system described above is simple to build, cleverly utilizing existing historical formula data to construct a LightGBM classification model for recommending color matching samples. This system efficiently groups and filters the formula data using both decision tree and LightGBM classification models, and integrates the color matching recommendation method into the color matching recommendation system for use by color matching operators. This makes the color matching process intelligent and automated, reducing reliance on human experience in formula management and recommendation processes, and improving the accuracy and efficiency of color matching formula recommendations.

[0019] Furthermore, the color matching classification module is used to obtain several color matching datasets based on the historical color matching formula database and a preset decision tree classification model, including: extracting material features and process features of different color matching formulas from the historical color matching formula database to obtain a material feature set and a process feature set; inputting the material feature set and the process feature set into the decision tree classification model so that the decision tree classification model splits the historical color matching formula database into several color matching datasets based on the material feature set and the process feature set; the color matching dataset is a dataset composed of historical color matching formula data with the same material features and the same process features.

[0020] Furthermore, the classification model training module is used to train a pre-constructed LightGBM initial classification model based on several color matching datasets to obtain a LightGBM classification model. This includes: preprocessing data based on the historical color matching formula database and dividing the dataset into a training set and a test set; training the pre-constructed LightGBM initial classification model based on the training set and specifying parameters to obtain a training model; evaluating the performance and tuning the parameters of the training model using a preset cross-validation method to obtain a validation model; and evaluating the validation model based on the test set to determine if the validation model meets preset training requirements, thus obtaining the LightGBM classification model.

[0021] Furthermore, the color matching recommendation module is used to obtain a color scheme that meets preset requirements based on the sample color value parameters and the target color system subset, including: calculating the Euclidean distance between the sample color value parameters and each color value in the target color system subset to obtain a distance value sequence; obtaining a target distance value that meets preset requirements from the distance value sequence, and using the target color system subset corresponding to the target distance value as the color scheme.

[0022] The above scheme is a specific implementation of the color matching recommendation method in a color matching recommendation system. Through the operation of modules such as color matching classification, classification model training, and color matching recommendation, operators can accurately and quickly obtain color matching recommendations by leveraging existing historical data of color matching formulas, thereby making the color matching process intelligent and automated, and reducing reliance on human experience. Attached Figure Description

[0023] Figure 1 This is a schematic flowchart of a color matching recommendation method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a color matching recommendation system architecture provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 This embodiment provides a color matching recommendation method, including the following steps: Step S1: Establish a historical database of color matching formulas; Step S2: Based on the historical color matching formula database and the preset decision tree classification model, obtain several color matching datasets; Step S3: Train the pre-built LightGBM initial classification model based on several of the color matching datasets to obtain the LightGBM classification model; Step S4: Obtain the sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched; Step S5: Based on the sample material parameters and sample process parameters, obtain a subset of the target formula database from the color matching formula history database; Step S6: Input the sample color value parameter into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched; Step S7: Obtain a subset of target color systems from the subset of the target formula database based on the sample color system; Step S8: Obtain a color scheme that meets the preset requirements based on the sample color value parameters and the target color system subset.

[0026] In this embodiment, a historical database of color matching formulas is first established by collecting existing color matching data and then filtered using a preset decision tree classification model to obtain a color matching dataset. This color matching data may include formula serial number, fabric material, dyeing process parameters (processing temperature, dye source, holding time), color value (RGB value), dye type, and concentration. Based on this, a pre-built LightGBM initial classification model is trained to obtain a LightGBM classification model, further dividing the color matching dataset into color system subsets. For a sample to be matched, its color value parameters, material parameters, and process parameters are first obtained. Based on the material and process parameters, a target formula database subset can be filtered from the historical color matching formula database. Then, the color parameters are input into the LightGBM classification model to determine the color system, and a target color system subset is obtained from the target formula database subset. Finally, a color matching scheme that meets preset requirements is obtained based on the color parameters and the target color system subset.

[0027] This embodiment digitizes the dyeing formula data and uses a decision tree classification model and a LightGBM classification model to efficiently group and filter the formula data, thereby making the color matching process intelligent and automated, reducing reliance on human experience, and improving the accuracy and efficiency of color matching formula recommendations.

[0028] Furthermore, the step of obtaining several color matching datasets based on the historical color matching formula database and the preset decision tree classification model includes: extracting material features and process features of different color matching formulas from the historical color matching formula database to obtain a material feature set and a process feature set; inputting the material feature set and the process feature set into the decision tree classification model, so that the decision tree classification model splits the historical color matching formula database into several color matching datasets S based on the material feature set and the process feature set. N (N=1,2,3,…); The color matching dataset is a dataset composed of historical color matching formulas with the same material characteristics and the same process characteristics.

[0029] In this embodiment, to obtain a color matching dataset composed of historical color matching formulas with the same material and process characteristics, a decision tree classification model is used. The model filters the historical color matching formula database according to requirements such as fabric material and dyeing process parameters. A decision tree is a recursive tree structure that classifies or regresses data through a series of rules. The decision tree classification model divides the historical color matching formula database into multiple color matching datasets by selecting the optimal split point, maximizing (or minimizing) the purity (or impurity) of each dataset. Its key construction principles include the following three parts: (1) Information gain: used for classification tasks. It calculates the information gain of a dataset after splitting it by a certain feature, and selects the feature with the largest information gain for splitting.

[0030] in, D For the dataset, A As a feature, v Features A The value of , D v In the feature A The subset of data with a value of v.

[0031] (2) Entropy: a measure of the purity of a dataset.

[0032] in p i For category i The proportion in the dataset.

[0033] (3) Gini Index: Used for classification tasks, the optimal split point is selected by calculating the Gini index.

[0034] By filtering through a preset decision tree classification model, the historical database of color matching formulas can be split into several color matching datasets based on the material feature set and the process feature set for subsequent construction of the lightGBM model.

[0035] Further, the step of training the pre-constructed LightGBM initial classification model based on the several color matching datasets to obtain the LightGBM classification model includes: preprocessing the data based on the historical color matching recipe database and dividing the dataset into a training set and a test set; training the pre-constructed LightGBM initial classification model based on the training set and specifying parameters to obtain a training model; evaluating the performance and tuning the parameters of the training model based on a preset cross-validation method to obtain a validation model; and evaluating the validation model based on the test set to determine that the validation model meets the preset training requirements, thus obtaining the LightGBM classification model.

[0036] Furthermore, if the verification model does not meet the preset training requirements, the verification model is retrained based on the training set and specified parameters to obtain a new training model and a new verification model, until the new verification model meets the preset training requirements.

[0037] This embodiment uses LightGBM to build a model for color classification. LightGBM is a distributed gradient boosting framework based on decision tree algorithms, characterized by high efficiency and accuracy, suitable for classification and regression tasks on large-scale data. LightGBM employs a histogram-based decision tree learning algorithm. It accelerates computation and reduces memory consumption by binning continuous data and constructing histograms. Unlike traditional layer-by-layer growth, LightGBM uses a leaf node growth strategy, selecting the optimal leaf node for growth at each split, thereby reducing the loss function more quickly. LightGBM reduces the computational complexity of feature splitting by discretizing continuous feature values ​​into histograms, while leveraging the cumulative properties of histograms to quickly find the optimal split point. Specifically, by constructing histograms, continuous feature values ​​are divided into multiple bins, each containing a segment of continuous values ​​as follows: in H j For the first j The cumulative value of each bucket, I This is an indicator function.

[0038] The goal of LightGBM is to optimize the loss function by iteratively fitting the residuals: in, y This is the actual value. F(x) For predicted values, l The loss function for a single sample.

[0039] Meanwhile, LightGBM supports multi-threading and GPU acceleration, enabling efficient model training through distributed learning. LightGBM can also handle large-scale class features, avoiding the high-dimensionality problem caused by one-hot encoding in traditional methods.

[0040] In this embodiment, the historical color formula database first needs to be preprocessed. This includes dividing the colors in the database into seven color systems (red, orange, yellow, green, cyan, blue, and purple) based on the CIELAB color model and OHTA color space. A pre-built LightGBM initial classification model is then established, and the dataset is divided into training and testing sets. Next, the pre-built LightGBM initial classification model is trained using the training set and specified parameters to obtain a training model. During training, the training model's performance is evaluated and parameters are fine-tuned using a pre-defined cross-validation method to obtain a validation model. The validation model is then evaluated using the test set to determine if it meets the pre-defined training requirements, thus obtaining the LightGBM classification model.

[0041] It should be noted that if the verification model does not meet the preset index requirements, the verification model should be retrained based on the training set and with specified parameters to obtain a new training model and a new verification model, until the new verification model meets the preset training requirements.

[0042] It should be noted that during model training, cross-validation is also needed to evaluate model performance and adjust hyperparameters to optimize model performance. The specific process is as follows: (1) Data Preparation: The data is divided into features and labels, and then divided into training and test sets using an 80 / 20 ratio. Finally, the features are standardized to ensure that different features have the same scale. The corresponding codes for the data labels (color classification) are as follows: (2) Define the model and parameter grid: Select the LightGBM classification model and list the hyperparameters that need to be adjusted and their value ranges.

[0043] (3) Cross-validation and grid search: Use GridSearchCV to perform hyperparameter search and cross-validation. Run grid search on the training set to find the optimal parameter combination.

[0044] (4) Evaluate the optimal model: Train the model using the optimal parameters: Train the final model based on the optimal parameters found by grid search, and evaluate the model performance using the test set.

[0045] It should be noted that after the LightGBM classification model was built, this model was used to further divide the database of the same material features and the same process features into 7 color system subsets S based on the color system. NK (N=1,2,3,…,K=1,2,…7). A suitable color matching dataset S is selected based on the sample material and process characteristics. N After measuring the color value of the sample, the target color system is obtained by inputting it into the model, and then a suitable subset S of the target color system is selected. NK .

[0046] Further, obtaining a color scheme that meets preset requirements based on the sample color value parameters and the target color subset includes: calculating the sample color value parameters and the target color subset S respectively. NK The Euclidean distance of each color value in the color system is used to obtain a distance value sequence; the target distance value that meets the preset requirements is obtained from the distance value sequence, and the target color system subset corresponding to the target distance value is used as the color scheme.

[0047] Further, the step of calculating the Euclidean distance between the sample color value parameter and the color value of each color system in the target color system subset to obtain a distance value sequence is specifically as follows: For any color system in the target color system subset, the Euclidean distance between its color system value and the sample color value parameter is calculated as follows: in ρ For Euclidean distance, R x , G x , B x The sample color value parameter, R n , G n , B n For any color system color value parameter in the target color system subset; The distance value sequence is a data sequence consisting of the distance values ​​between each color system color value and the sample color value parameter.

[0048] The Euclidean distance obtained here is used to characterize the color difference between the sample and each color system in the target color system subset. It should be noted that by sorting the sequence of distance values ​​obtained by traversing the sample color values ​​and the target color system subset from smallest to largest, the dyeing formulas corresponding to the first few distance values ​​can be selected as the proposed recommended formula output. The output formula data includes the dye type and concentration value.

[0049] Please see Figure 2 This embodiment also provides a color matching recommendation system, including: a database construction module for establishing a historical database of color matching recipes; a color matching classification module for obtaining several color matching datasets based on the historical database of color matching recipes and a preset decision tree classification model; a classification model training module for training a pre-constructed LightGBM initial classification model based on the several color matching datasets to obtain a LightGBM classification model; a sample parameter acquisition module for acquiring sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched; a target recipe screening module for acquiring a subset of target recipe databases from the historical database of color matching recipes based on the sample material parameters and sample process parameters; a sample color system acquisition module for inputting the sample color value parameters into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched; a target color system acquisition module for acquiring a subset of target color systems from the subset of target recipe databases based on the sample color systems; and a color matching recommendation module for acquiring color matching schemes that meet preset requirements based on the sample color value parameters and the subset of target color systems.

[0050] The system provided in this embodiment is simple to construct. It cleverly utilizes existing historical formula data to build a LightGBM classification model for recommending color matching samples. The system efficiently groups and filters the formula data using a decision tree classification model and a LightGBM classification model, and integrates the color matching recommendation method into a cloud platform system for use by color matching operators. This improves the accuracy and efficiency of color matching formula recommendations, ensuring the stability and consistency of dyeing effects.

[0051] Furthermore, the color matching classification module is used to obtain several color matching datasets based on the historical color matching formula database and a preset decision tree classification model, including: extracting material features and process features of different color matching formulas from the historical color matching formula database to obtain a material feature set and a process feature set; inputting the material feature set and the process feature set into the decision tree classification model so that the decision tree classification model splits the historical color matching formula database into several color matching datasets based on the material feature set and the process feature set; the color matching dataset is a dataset composed of historical color matching formula data with the same material features and the same process features.

[0052] Furthermore, the classification model training module is used to train a pre-constructed LightGBM initial classification model based on several color matching datasets to obtain a LightGBM classification model. This includes: preprocessing data based on the historical color matching formula database and dividing the dataset into a training set and a test set; training the pre-constructed LightGBM initial classification model based on the training set and specifying parameters to obtain a training model; evaluating the performance and tuning the parameters of the training model using a preset cross-validation method to obtain a validation model; and evaluating the validation model based on the test set to determine if the validation model meets preset training requirements, thus obtaining the LightGBM classification model.

[0053] Furthermore, the color matching recommendation module is used to obtain a color scheme that meets preset requirements based on the sample color value parameters and the target color system subset, including: calculating the Euclidean distance between the sample color value parameters and each color value in the target color system subset to obtain a distance value sequence; obtaining a target distance value that meets preset requirements from the distance value sequence, and using the target color system subset corresponding to the target distance value as the color scheme.

[0054] This embodiment is a specific implementation of the color matching recommendation method on a color matching recommendation system. Through the operation of modules such as color matching classification, classification model training, and color matching recommendation, operators can accurately and quickly obtain color matching recommendations by leveraging existing historical data of color matching formulas. This makes the color matching process intelligent and automated, improves data management efficiency, and reduces reliance on human experience.

[0055] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A color matching recommendation method, characterized in that, Includes the following steps: Establish a historical database of color matching formulas; Based on the historical color formula database and a preset decision tree classification model, several color formula datasets are obtained, including: extracting material features and process features of different color formulas from the historical color formula database to obtain material feature sets and process feature sets; inputting the material feature sets and process feature sets into the decision tree classification model, so that the decision tree classification model splits the historical color formula database into several color formula datasets based on the material feature sets and process feature sets; the color formula datasets are datasets composed of historical color formula data with the same material features and the same process features; The pre-constructed LightGBM initial classification model is trained based on several of the aforementioned color datasets to obtain the LightGBM classification model; Obtain the sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched; Based on the sample material parameters and sample process parameters, a subset of the target formula database is obtained from the historical color formula database. This includes: after the LightGBM classification model is built, the database of the same sample material parameters and the same sample process parameters is divided into 7 color system subsets according to the color system using the LightGBM classification model; and suitable color matching datasets are selected from the 7 color system subsets according to the sample material parameters and sample process parameters to obtain the subset of the target formula database. Input the sample color value parameter into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched; Based on the sample color system, obtain a subset of the target color system from the subset of the target formula database; Based on the sample color value parameters and the target color system subset, a color scheme that meets the preset requirements is obtained, including: calculating the Euclidean distance between the sample color value parameters and each color value in the target color system subset to obtain a distance value sequence; obtaining a target distance value that meets the preset requirements from the distance value sequence, and using the target color system subset corresponding to the target distance value as the color scheme.

2. The color matching recommendation method according to claim 1, characterized in that, The process of training a pre-built initial LightGBM classification model based on the aforementioned color matching datasets to obtain a LightGBM classification model includes: Data preprocessing is performed on the aforementioned color scheme datasets, and the preprocessed color scheme datasets are divided into training sets and test sets. The pre-built LightGBM initial classification model is trained based on the training set and specified parameters to obtain the trained model; The training model is evaluated and its parameters are tuned based on a pre-defined cross-validation method to obtain a validation model. The validation model is evaluated based on the test set. If the validation model meets the preset training requirements, the LightGBM classification model is obtained.

3. The color matching recommendation method according to claim 2, characterized in that, The step of evaluating the validation model based on the test set to determine if the validation model meets the preset training requirements, thus obtaining the LightGBM classification model, further includes: If the verification model does not meet the preset training requirements, the verification model is retrained based on the training set and specified parameters to obtain a new training model and a new verification model, until the new verification model meets the preset training requirements.

4. The color matching recommendation method according to claim 1, characterized in that, The step of calculating the Euclidean distance between the sample color value parameter and the color value of each color system in the target color system subset to obtain a distance value sequence includes: For any color system in the target color system subset, the Euclidean distance between its color system value and the sample color value parameter is calculated as follows: in ρ For Euclidean distance, R x , G x , B x The sample color value parameter, R n , G n , B n For any color system color value parameter in the target color system subset; The distance value sequence is a data sequence consisting of the distance values ​​between each color system color value and the sample color value parameter.

5. A color matching recommendation system, characterized in that, For performing a color matching recommendation method as described in any one of claims 1 to 4, the system comprises: The database building module is used to create a historical database of color matching recipes; The color matching classification module is used to obtain several color matching datasets based on the historical color matching formula database and a preset decision tree classification model. This includes: extracting material features and process features of different color matching formulas from the historical color matching formula database to obtain a material feature set and a process feature set; inputting the material feature set and the process feature set into the decision tree classification model, so that the decision tree classification model splits the historical color matching formula database into several color matching datasets based on the material feature set and the process feature set; each color matching dataset is a dataset composed of historical color matching formula data with the same material features and the same process features. The classification model training module is used to train the pre-built LightGBM initial classification model based on several of the color matching datasets to obtain the LightGBM classification model. The sample parameter acquisition module is used to acquire sample color value parameters, sample material parameters, and sample process parameters based on the sample to be matched. The target formula screening module is used to obtain a subset of the target formula database from the historical color formula database based on the sample material parameters and sample process parameters. The module includes: after the LightGBM classification model is built, the database of the same sample material parameters and the same sample process parameters is divided into 7 color system subsets according to the color system using the LightGBM classification model; and the appropriate color matching dataset is selected from the 7 color system subsets according to the sample material parameters and sample process parameters to obtain the subset of the target formula database. The sample color system acquisition module is used to input the sample color value parameter into the LightGBM classification model, so that the LightGBM classification model outputs the sample color system corresponding to the sample to be matched. The target color system acquisition module is used to acquire a subset of target color systems from the target formula database subset based on the sample color system; The color matching recommendation module is used to obtain a color scheme that meets preset requirements based on the sample color value parameters and the target color system subset, including: calculating the Euclidean distance between the sample color value parameters and the color values ​​of each color system in the target color system subset to obtain a distance value sequence; obtaining a target distance value that meets preset requirements from the distance value sequence, and using the target color system subset corresponding to the target distance value as the color scheme.

6. A color matching recommendation system according to claim 5, characterized in that, The classification model training module is used to train a pre-built LightGBM initial classification model based on several color matching datasets to obtain a LightGBM classification model, including: Data preprocessing is performed on the aforementioned color scheme datasets, and the preprocessed color scheme datasets are divided into training sets and test sets. The pre-built LightGBM initial classification model is trained based on the training set and specified parameters to obtain the trained model; The training model is evaluated and its parameters are tuned based on a pre-defined cross-validation method to obtain a validation model. The validation model is evaluated based on the test set. If the validation model meets the preset training requirements, the LightGBM classification model is obtained.

Images