COB lamp bead color calibration classification method and system based on unsupervised machine learning and K-means clustering algorithm

Abstract

The invention discloses a COB lamp bead color calibration classification method based on unsupervised machine learning and a K-means clustering algorithm. The COB lamp bead color calibration classification method comprises the following steps: acquiring actual light color characteristics of COB lamp beads to be classified; and inputting the actual light color characteristics into a pre-trained lamp bead classification model for classification, wherein the lamp bead classification model is obtained through training in an unsupervised machine learning mode based on a K-means clustering algorithm. The invention provides a COB lamp bead color calibration classification method and system based on unsupervised machine learning and a K-means clustering algorithm, the COB lamp beads can be classified based on the multi-dimensional light color characteristics of the COB lamp beads, high-efficiency and high-precision light color calibration can be performed on the COB lamp beads based on the types of the lamp beads after classification is completed, and the accuracy of the COB lamp beads is improved. And the COB lamp bead can fully meet the use requirements.

Description

Technical Field

[0001] The present invention relates to the field of COB lamp beads, and more particularly to a COB lamp bead light color calibration and classification method and system based on unsupervised machine learning and K-means clustering algorithm. Background Art

[0002] In professional film and television fill lights, COB lamps are widely used due to their advantages such as high brightness and excellent color rendering. However, due to manufacturing processes and other factors, different COB lamps have significant differences in light color characteristics, such as luminous flux, color temperature, color rendering index, and single-channel saturation. This makes it almost impossible to perform independent light color calibration for each COB lamp in large quantities in practical applications. Calibration of COB lamps in batches also suffers from poor calibration accuracy, making it difficult for COB lamps to meet the strict light color consistency and accuracy requirements of professional film and television shooting. Summary of the Invention

[0003] In order to address the deficiencies in the prior art, the present invention provides a COB lamp bead light color calibration and classification method and system based on unsupervised machine learning and K-means clustering algorithm, which can classify COB lamp beads based on their multi-dimensional light color characteristics. After the classification is completed, the COB lamp beads can be calibrated with high efficiency and high precision based on the lamp bead category, so that the COB lamp beads can fully meet the usage requirements.

[0004] In order to achieve the above object, the specific scheme adopted by the present invention is:

[0005] A COB lamp pearlescent color calibration classification method based on unsupervised machine learning and K-means clustering algorithm includes the following steps:

[0006] Obtain the actual light color characteristics of the COB lamp beads to be classified;

[0007] The actual characteristics of light color are input into the pre-trained lamp bead classification model for classification. The lamp bead classification model is trained by unsupervised machine learning based on the K-means clustering algorithm. The method of training the lamp bead classification model includes:

[0008] Constructing light color characteristic vectors of multiple COB (Chip On Board) lamp beads, where the light color characteristic vectors include multiple light color characteristics;

[0009] Select at least two light color characteristic vectors as initial cluster centers;

[0010] Calculate the Euclidean distance between the remaining light color characteristic vectors and the initial cluster center, and perform preliminary classification on the remaining light color characteristic vectors based on the Euclidean distance to obtain multiple lamp bead categories;

[0011] Update the initial cluster center based on the lamp bead category to obtain the optimized cluster center;

[0012] Calculate the intra-cluster sum of squared errors for each lamp bead category;

[0013] Increase the number of initial cluster centers to update the sum of squared errors within the cluster;

[0014] When the sum of squared errors within a cluster is stable, multiple optimal cluster centers are obtained.

[0015] Preferably, the light color characteristic vector of the COB lamp bead is

[0016]

[0017] Among them, CL is cool white illuminance, WL is warm white illuminance, CT is cool white color temperature, WT is warm white color temperature, CX is cool white color rendering index, WX is warm white color rendering index, Cduv is cool white color deviation, Wduv is warm white color deviation, RL is red light illuminance, GL is green light illuminance, BL is blue light illuminance, RS is red light saturation, GS is green light saturation, BS is blue light saturation, RH is red hue, GH is green hue, and BH is blue hue.

[0018] Preferably, the method of selecting at least two light color characteristic vectors as initial cluster centers includes:

[0019] Calculate the spatial distribution density of all light color characteristic vectors;

[0020] Determine several spatial regions based on the calculation results of spatial distribution density;

[0021] In each spatial region, a light color characteristic vector is randomly selected as the initial cluster center;

[0022] The initial cluster center is expressed as

[0023]

[0024] Wherein, j is the serial number of the light color characteristic vector.

[0025] Preferably, the method for calculating the Euclidean distances between the remaining light color characteristic vectors and the initial cluster center includes:

[0026] Calculate the light color characteristic vector C i With the initial cluster center C j The difference in each light color characteristic is CL i -CL j , WL i -WLj , CT i -CT j , WT i -WT j , CX i -CX j , WX i -WX j , Cduv i -Cduv j , Wduv i -Wduv j , RL i -RL j , GL i -GL j , BL i -BL j , RS i -RS j , GS i -GS j , B.S. i -BS j , RH i -RH j , GH i -GH j and BH i -BH j ;

[0027] Calculate the square of each difference, respectively (CL i -CL j ) 2 , (WL i -WL j ) 2 , (CT i -CT j ) 2 , (WT i -WT j ) 2 , (CX i -CX j ) 2 ,(WX i -WX j ) 2 , (Cduv i -Cduv j ) 2 , (Wduv i -Wduv j ) 2 , (RL i -RL j ) 2 , (GL i -GL j) 2 , (BL i -BL j ) 2 , (RS i -RS j ) 2 , (GS i -GS j ) 2 , (BS i -BS j ) 2 , (RH i -RH j ) 2 , (GH i -GH j ) 2 and (BH i -BH j ) 2 ;

[0028] Calculate the light color characteristic vector C i With the initial cluster center C j The overall difference

[0029]

[0030] Calculate the light color characteristic vector C i With the initial cluster center C j Euclidean distance

[0031] Preferably, the method for updating the initial cluster center based on the lamp bead category to obtain the optimized cluster center includes:

[0032] Calculate the mean value of each light color characteristic in all light color characteristic vectors within the lamp bead category;

[0033] Optimize cluster centers based on all mean components.

[0034] Preferably, the sum of squares of the intra-cluster errors of each lamp bead category is calculated, and the cumulative sum of the sums of squares of all intra-cluster errors is calculated and recorded as the total intra-cluster error;

[0035] Whether the intra-cluster sum of squared errors is stable is determined based on the degree of change in the sum of intra-cluster errors before and after the number of initial cluster centers increases.

[0036] A COB lamp bead classification system is used to implement the above-mentioned COB lamp bead color calibration classification method based on unsupervised machine learning and K-means clustering algorithm. The system includes:

[0037] A data generation module is used to construct a light color characteristic vector of multiple COB lamp beads, where the light color characteristic vector includes multiple light color characteristics;

[0038] A category building module is used to obtain multiple optimal cluster centers based on light color feature vectors;

[0039] The classification module is used to classify the COB lamp beads to be classified based on the optimal cluster center.

[0040] Based on the K-means algorithm, this invention uses unsupervised learning to train a lamp bead classification model capable of accurately classifying COB lamp beads. This trained lamp bead classification model can then be used to classify COB lamp beads to be processed. COB lamp beads classified into the same lamp bead category have similar light color characteristics. Therefore, after the classification is completed, the COB lamp beads can be calibrated with high precision and efficiency based on the classification results, ensuring that the COB lamp beads fully meet the user's needs.

[0041] The present invention effectively solves the problem in the prior art of being unable to determine how many categories COB lamp beads should be divided into, and how to accurately classify COB lamp beads incoming in batches. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0043] Figure 1 This is a flowchart of the training method of the lamp bead classification model;

[0044] Figure 2 It is a structural block diagram of the COB lamp bead classification system of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0046] A pearlescent color calibration classification method for COB lamps, including T1 to T2, based on unsupervised machine learning and K-means clustering algorithm.

[0047] T1. Obtain the actual light color characteristics of the COB lamp beads to be classified.

[0048] T2. Input the actual characteristics of light color into the pre-trained lamp bead classification model for classification.

[0049] The lamp bead classification model is trained by unsupervised machine learning based on the K-means clustering algorithm, and the method for training the lamp bead classification model includes S1 to S8.

[0050] S1. Construct the light color characteristic vectors of multiple COB (Chip On Board) lamp beads. The light color characteristic vectors include multiple light color characteristics. More specifically, the light color characteristic vectors of COB lamp beads are Among them, CL is the cool white illuminance, the unit is lux, which is used to measure the cool white light output performance of COB lamp beads, and can reflect the light intensity when the cool white light emitted by COB lamp beads hits the surface of the object; WL is the warm white illuminance, the unit is lux, which is used to measure the warm white light output performance of COB lamp beads, and can reflect the irradiation intensity when the warm white light emitted by COB lamp beads hits the surface of the object; CT is the cool white temperature, the unit is Kelvin, which is used to describe the color characteristics of the cool white light of COB lamp beads. The cool white temperature determines the cool white light hue tendency of COB lamp beads and affects the visual effect of objects under cool white light; WT The unit is Kelvin, which is used to describe the color characteristics of the warm white light of COB lamp beads. The warm white temperature determines the hue tendency of the warm white light of COB lamp beads and affects the visual effect of objects under warm white light. CX is the cool white color rendering index, which ranges from 0 to 100 and is used to reflect the ability of the cool white light emitted by COB lamp beads to restore the color of objects. The higher the value of the cool white color rendering index, the more realistic the color restoration. WX is the warm white color rendering index, which ranges from 0 to 100 and is used to reflect the ability of the warm white light emitted by COB lamp beads to restore the color of objects. Cduv is the cool white color deviation, which is used to measure the cool white color rendering index of COB lamp beads. The degree of deviation between the color light and the color of the Planckian radiator, and reflects the chromaticity accuracy of the cold white light; Wduv is the warm white color deviation, which is used to measure the degree of deviation between the warm white light of the COB lamp bead and the color of the Planckian radiator, and reflects the chromaticity quality of the warm white light of the OB lamp bead; RL is the red light illuminance, the unit is lux, which is used to indicate the irradiation intensity of the red light emitted by the COB lamp bead; GL is the green light illuminance, the unit is lux, which is used to indicate the irradiation intensity of the blue light emitted by the COB lamp bead; BL is the blue light illuminance, the unit is lux, which is used to indicate the irradiation intensity of the green light emitted by the COB lamp bead; RS is the red light saturation, which is used to indicate the C OB lamp beads represent the vividness or purity of the red light; GS is the green light saturation, which is used to indicate the vividness or purity of the blue light emitted by COB lamp beads; BS is the blue light saturation, which is used to indicate the vividness or purity of the green light emitted by COB lamp beads; RH is the red hue, which describes the position of the red light emitted by COB lamp beads on the hue circle and determines the specific hue of red; GH is the green hue, which describes the position of the green light emitted by COB lamp beads on the hue circle and determines the specific hue of green; BH is the blue hue, which describes the position of the blue light emitted by COB lamp beads on the hue circle and determines the specific hue of blue. It should be noted that the actual light color characteristics of the COB lamp beads to be classified are also represented in the form of light color characteristic vectors.

[0051] Each light color characteristic in the above light color characteristic vector is a conventional parameter of COB lamp beads.

[0052] S2. Select at least two light color characteristic vectors as initial cluster centers. The initial cluster centers can be randomly selected from all light color characteristic vectors. However, to make the initial cluster centers more representative, in one embodiment of the present invention, the method of selecting at least two light color characteristic vectors as initial cluster centers includes S21 to S23.

[0053] S21. Calculate the spatial distribution density of all light color characteristic vectors. Accordingly, a multidimensional space must be defined. Because the light color characteristic vectors contain a total of seventeen light color characteristics, in this embodiment, a seventeen-dimensional multidimensional space is also pre-set. All light color characteristic vectors are then represented in this multidimensional space. Each light color characteristic vector corresponds to a point in the multidimensional space, which can be called a characteristic point. The spatial distribution density of the light color characteristic vectors can be determined based on the position of the characteristic point.

[0054] S22. Determine a plurality of spatial regions based on the calculation result of the spatial distribution density. Specifically, after obtaining the spatial distribution density of the light color characteristic vectors, a portion of the region where multiple light color characteristic vectors are distributed can be considered as a spatial region. That is, each spatial region contains characteristic points of multiple light color characteristic vectors, and the distance between any two characteristic points in the same spatial region does not exceed a preset distance threshold.

[0055] S23. Randomly select a light color characteristic vector in each spatial region as the initial cluster center. The initial cluster center is expressed as Where j is the sequence number of the light color characteristic vector. Because the distances between the characteristic points in each spatial region are relatively close, the similarity of the corresponding light color characteristic vectors is also high. The initial cluster center selected based on this method can be close to the remaining light color characteristic vectors, and the differences between the selected initial cluster centers are avoided to be too small, thereby accelerating the convergence of subsequent processes.

[0056] S3. Calculate the Euclidean distance between the remaining light color characteristic vectors and the initial cluster center, and preliminarily classify the remaining light color characteristic vectors based on the Euclidean distance to obtain multiple lamp bead categories. More specifically, the method for calculating the Euclidean distance between the remaining light color characteristic vectors and the initial cluster center includes S31 to S34.

[0057] S31. Calculate light color characteristic vector C i With the initial cluster center C j The difference in each light color characteristic is CL i -CL j , WL i -WL j , CT i-CT j , WT i -WT j , CX i -CX j , WX i -WX j , Cduv i -Cduv j , Wduv i -Wduv j , RL i -RL j , GL i -GL j , BL i -BL j , RS i -RS j , GS i -GS j , B.S. i -BS j , RH i -RH j , GH i -GH j and BH i -BH j These differences reflect the light color characteristic vector C i With the initial cluster center C j The degree of difference in the characteristics of various light colors.

[0058] S32, calculate the square of each difference, respectively (CL i -CL j ) 2 , (WL i -WL j ) 2 , (CT i -CT j ) 2 , (WT i -WT j ) 2 , (CX i -CX j ) 2 ,(WX i -WX j ) 2 , (Cduv i -Cduv j ) 2 , (Wduv i -Wduv j ) 2 , (RL i -RL j )2 , (GL i -GL j ) 2 , (BL i -BL j ) 2 , (RS i -RS j ) 2 , (GS i -GS j ) 2 , (BS i -BS j ) 2 , (RH i -RH j ) 2 , (GH i -GH j ) 2 and (BH i -BH j ) 2 By squaring the difference, the positive and negative effects of the difference can be eliminated, and the effect of the larger difference can be amplified in the subsequent process.

[0059] S33, calculate the light color characteristic vector C i With the initial cluster center C j The overall difference The overall difference reflects the light color characteristic vector C i With the initial cluster center C j The overall degree of variability in all light color characteristics.

[0060] S34. Calculate light color characteristic vector C i With the initial cluster center C j Euclidean distance Euclidean distance geometrically represents the light color characteristic vector C i With the initial cluster center C j In the above multidimensional space, the smaller the straight-line distance is, the greater the light color characteristic vector C is. i With the initial cluster center C j The more similar they are in all light color properties.

[0061] Furthermore, after calculating each light color characteristic vector C i With each initial cluster center C j After the Euclidean distance between the initial cluster center C j Based on the Euclidean distance, the light color characteristic vector C iFor example, if the light color characteristic vectors C1, C2, and C3 are all the same as the initial cluster center C 15 Similarly, the light color characteristic vectors C1, C2, C3, C 15 Combined into one lamp bead category.

[0062] S4, based on the lamp bead category, the initial cluster center is updated to obtain the optimized cluster center. After obtaining multiple lamp bead categories, although each lamp bead category corresponds to multiple light color characteristic vectors C i With an initial cluster center C j In the light color characteristic vector C i Both are the same as the initial cluster center C j Similar, but the degree of similarity is still different, there may be a light color characteristic vector C i With the initial cluster center C j The similarity is significantly higher than that of other light color characteristic vectors C i With the initial cluster center C j Therefore, although it is possible to use the initial cluster center C j Characterizing the lamp bead category, but there is still the possibility of poor accuracy. To solve this problem, the present invention further updates the initial cluster center. Specifically, the method of updating the initial cluster center based on the lamp bead category to obtain the optimized cluster center includes S41 to S42.

[0063] S41, calculate the mean value of each light color characteristic in all light color characteristic vectors within the lamp bead category. For example, if a lamp bead category S j There is N=|S j | light color characteristic vectors, respectively denoted as And there are Where l = 1, 2, ^, N. The calculation methods for the mean values of the various light color characteristics are as follows:

[0064] S42. Optimize cluster centers based on all mean values. In this way, the optimized cluster centers can be recalculated based on the dimensional values of all light color characteristic vectors in the current lamp bead category. That is, the optimized cluster centers are recalculated based on the light color characteristics in all light color characteristic vectors, so that the optimized cluster centers can better represent the light color characteristics of the lamp bead category.

[0065] Although after optimization, the similarity between the optimized cluster center and other light color characteristic vectors in the corresponding lamp bead category is higher than that between the initial cluster center, there are still cases where the optimized cluster center and the light color characteristic vector differ too much. The main reason is that if the number of initial cluster centers is small, the number of optimized cluster centers is also small. This may result in some light color characteristic vectors that differ significantly from the optimized cluster center being included in the lamp bead category corresponding to the optimized cluster center. To solve this problem, the present invention further optimizes the number of optimized cluster centers. The specific method is as follows.

[0066] S5. Calculate the sum of squared errors within each cluster of each lamp bead category. The optimized cluster center in the lamp bead category is recorded as The remaining light color characteristic vectors are recorded as Then the sum of squares of the intra-cluster errors of each lamp bead category is

[0067] Furthermore, the sum of squared errors within each cluster is calculated for each lamp bead category, and the sum of the sum of squared errors within all clusters is calculated and recorded as the total sum of squared errors within the cluster. Where k is the initial cluster center, optimized cluster center and the number of lamp bead categories.

[0068] S6. Increase the number of initial cluster centers to update the intra-cluster error sum of squares. Specifically, after obtaining the intra-cluster error sum corresponding to a k value, the k value is expanded, and then the above process is repeated to generate a new intra-cluster error sum.

[0069] S7. When the intra-cluster sum of squared errors is stable, multiple optimal cluster centers are obtained. Taking into account that the number of intra-cluster sums of squared errors will gradually increase with the increase of k value, in order to facilitate the determination of whether the intra-cluster sum of squared errors tends to be stable, the stability of the intra-cluster sum of squared errors is determined based on the degree of change in the sum of intra-cluster errors before and after the initial number of cluster centers increases. More specifically, as K increases, SSE usually decreases gradually, because more lamp bead categories mean that the light color characteristics of COB lamp beads in each lamp bead category will be more similar, and accordingly, the intra-cluster sum of squared errors will decrease. When k is small, increasing k will significantly reduce SSE; but when k increases to a certain extent, the effect of continuing to increase k on reducing SSE is no longer obvious. At this time, the curve will have an inflection point. The k value corresponding to this inflection point is the best, and the corresponding k optimized cluster centers are the optimal cluster centers.

[0070] S8. Classify the COB lamp beads to be classified based on the optimal cluster center. Specifically, for the COB lamp beads to be classified, their actual light color characteristics can be measured and a light color characteristic vector can be formed. Then, the difference between the light color characteristics and the optimal cluster center is calculated. If the difference is lower than a preset threshold, the COB lamp beads to be classified can be classified into the lamp bead category corresponding to the optimal cluster center.

[0071] The present invention's lamp bead light color calibration and classification method constructs a lamp bead classification model based on the K-means classification algorithm. This model can be trained without manual labeling, using an unsupervised learning approach. The trained lamp bead classification model can classify pending COB lamp beads into multiple lamp bead categories, with COB lamp beads classified into the same lamp bead category having similar light color characteristics. Therefore, after classification is complete, the COB lamp beads can be calibrated with high precision and efficiency based on the classification results, ensuring that the COB lamp beads fully meet user requirements.

[0072] Furthermore, after the COB lamp beads are classified, the light color of the COB lamp beads can be calibrated according to the light color characteristics of the optimal cluster center. The details are as follows.

[0073] The above-mentioned COB lamp bead classification method is used to determine multiple optimal cluster centers, and the COB lamp beads to be classified are classified based on the optimal cluster centers.

[0074] The light color of the classified COB lamp beads is calibrated based on the preset standard color calibration data. The standard color calibration data corresponds one-to-one with the optimal cluster center. After the optimal cluster center is determined, the standard color calibration data is generated based on the difference between the light color characteristics of the optimal cluster center and the standard light color characteristics. After the COB lamp beads are classified into the lamp bead category where the optimal cluster center is located, they can be calibrated based on the standard color calibration data, achieving high-efficiency and high-precision calibration of the COB lamp beads, thereby ensuring that the COB lamp beads can fully meet the needs.

[0075] The present invention also provides a COB lamp bead classification system for implementing the above-mentioned COB lamp bead classification method, and the system includes a data generation module, a category construction module and a classification module.

[0076] The data generation module is used to construct the light color characteristic vector of multiple COB lamp beads, and the light color characteristic vector includes multiple light color characteristics.

[0077] The category building module is used to obtain multiple optimal clustering centers based on the light color feature vector.

[0078] The classification module is used to classify the COB lamp beads to be classified based on the optimal cluster center. Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the system and device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. In the several embodiments provided in the present disclosure, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0079] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules may be selected to achieve the purpose of the present embodiment according to actual needs.

[0080] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0081] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A COB lamp pearlescent color calibration classification method based on unsupervised machine learning and K-means clustering algorithm, characterized in that: The steps include: Obtain the actual light color characteristics of the COB lamp beads to be classified; The actual characteristics of light color are input into the pre-trained lamp bead classification model for classification. The lamp bead classification model is trained by unsupervised machine learning based on the K-means clustering algorithm. The method of training the lamp bead classification model includes: Constructing light color characteristic vectors of multiple COB (Chip On Board) lamp beads, where the light color characteristic vectors include multiple light color characteristics; Select at least two light color characteristic vectors as initial cluster centers; Calculate the Euclidean distance between the remaining light color characteristic vectors and the initial cluster center, and perform preliminary classification on the remaining light color characteristic vectors based on the Euclidean distance to obtain multiple lamp bead categories; Update the initial cluster center based on the lamp bead category to obtain the optimized cluster center; Calculate the sum of squared errors within each cluster for each lamp bead category; Increase the number of initial cluster centers to update the sum of squared errors within the cluster; When the sum of squared errors within a cluster is stable, multiple optimal cluster centers are obtained.

2. A COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as claimed in claim 1, characterized in that: The light color characteristic vector of COB lamp beads is: Among them, CL is cool white illuminance, WL is warm white illuminance, CT is cool white color temperature, WT is warm white color temperature, CX is cool white color rendering index, WX is warm white color rendering index, Cduv is cool white color deviation, Wduv is warm white color deviation, RL is red light illuminance, GL is green light illuminance, BL is blue light illuminance, RS is red light saturation, GS is green light saturation, BS is blue light saturation, RH is red hue, GH is green hue, and BH is blue hue.

3. The COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as described in claim 2, characterized in that: The method of selecting at least two light color characteristic vectors as initial cluster centers includes: Calculate the spatial distribution density of all light color characteristic vectors; Determine several spatial regions based on the calculation results of spatial distribution density; In each spatial region, a light color characteristic vector is randomly selected as the initial cluster center; The initial cluster center is expressed as Wherein, j is the serial number of the light color characteristic vector.

4. The COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as described in claim 3, characterized in that: Methods for calculating the Euclidean distances between the remaining light color characteristic vectors and the initial cluster center include: Calculate the light color characteristic vector C i With the initial cluster center C j The differences in each light color characteristic are CLi-CLj, WLi-WLj, CTi-CTj, WTi-WTj, CXi-CXj, WXi-WXj, Cduvi-Cduvj, Wduvi-Wduvj, RLi-RLj, GLi-GLj, BLi-BLj, RSi-RSj, GSi-GSj, BSi-BSj, RHi-RHj, GHi-GHj and BHi-BHj; Calculate the square of each difference, respectively, (CLi-CLj)2, (WLi-WLj)2, (CTi-CTj)2, (WTi-WTj)2, (CXi-CXj)2, (WXi-WXj)2, (Cduvi-Cduvj)2, (Wduvi-Wduvj)2, (RLi-RLj)2, (GLi-GLj)2, (BLi-BLj)2, (RSi-RSj)2, (GSi-GSj)2, (BSi-BSj)2, (RHi-RHj)2, (GHi-GHj)2 and (BHi-BHj)2; Calculate the light color characteristic vector C i With the initial cluster center C j The overall difference Calculate the light color characteristic vector C i With the initial cluster center C j Euclidean distance 5. The COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as claimed in claim 1, characterized in that: Methods for updating the initial cluster center based on the lamp bead category to obtain the optimized cluster center include: Calculate the mean value of each light color characteristic in all light color characteristic vectors within the lamp bead category; Optimize cluster centers based on all mean components.

6. The COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as claimed in claim 1, characterized in that: Calculate the intra-cluster error sum of squares for each lamp bead category, and calculate the cumulative sum of all intra-cluster error sums as the total intra-cluster error; Whether the intra-cluster sum of squared errors is stable is determined based on the degree of change in the sum of intra-cluster errors before and after the number of initial cluster centers increases.

7. COB lamp bead classification system, characterized by: A system for implementing a COB lamp pearlescent color calibration and classification method based on unsupervised machine learning and K-means clustering algorithm as described in any one of claims 1 to 6, the system comprising: A data generation module is used to construct a light color characteristic vector of multiple COB lamp beads, where the light color characteristic vector includes multiple light color characteristics; A category building module is used to obtain multiple optimal cluster centers based on light color feature vectors; The classification module is used to classify the COB lamp beads to be classified based on the optimal cluster center.

Images