Research On Prediction And Correction Methods Of RGB-LED Chromaticity Parameters

With the continuous development of science and technology and people's continuous pursuit of quality of life,RGB-LEDs play an important role in many fields,among which color consistency plays a vital role.Due to factors such as manufacturing process and junction temperature,different RGB-LEDs will have a certain color difference when they are lit with the same R,G,and B duty cycles.The current color difference detection methods for RGB-LEDs are too low in efficiency,and the accuracy of color difference calibration is not high enough.Therefore,this paper proposes a method for intelligent prediction and calibration of RGB-LED chromaticity parameters.The main work and innovations of this article have the following two parts:The first is the color difference prediction part.By combining the chromaticity parameters of RGB-LEDs with machine learning algorithms,intelligent predictions can be made on the color difference when RGB-LEDs light up different colors.This article uses common standards in the lighting industry,and the chromaticity tolerance setting value is0.0055.First,combining the ideas of random forest and XGBoost,this paper designs a classification algorithm and named it the random extreme gradient model(RF-XGBoost).Then,based on the combination of chromaticity related theoretical analysis and color difference analysis of experimental results,nine characteristic parameters that affect chromatic aberration are proposed.This paper compares the color difference prediction results of RGB-LED with decision tree,random forest,XGBoost,SVM and RF-XGBoost,and verifies the superiority of RF-XGBoost algorithm in color difference prediction.The second is the color correction part.In this paper,by analyzing the causes of RGB-LED color difference,combining the theoretical basis of colorimetry and actual calibration data analysis,a high-precision color difference calibration algorithm is proposed and named as the high-precision color calibration algorithm(Cor-Calibration).By comparing the uncalibrated color difference of the RGB-LED,the color difference of the initial calibration algorithm and the color difference of the Cor-Calibration algorithm,the results show that the Cor-Calibration algorithm has obvious advantages for high-precision calibration of color difference,and significantly improves the pass rate of color difference.This paper builds an experimental verification platform through the server,the lower computer and the upper computer to verify the effectiveness of the color difference prediction and calibration methods.First,the color difference prediction model is deployed in the server,and then the color difference calibration part is deployed in the lower computer KEA64.Through the coordinated control of the upper computer,the server,and the lower computer,the prediction and calibration results are finally displayed on the upper computer.In the analysis of the results of the prediction model,this paper uses a ten-fold cross-validation method to use False positiveed detection rate,recall rate,accuracy rate,and F₁?score as the evaluation indicators of the color difference prediction model.Experiments show that when the accuracy and F₁?score are not much different,the False positiveed detection rate and recall rate of RF-XGBoost are the best to reach 12.60%and 95.85%,respectively.In the analysis of the results of the color calibration model,this paper uses whether the x and y chromaticity coordinates of CIE1931-RGB are both less than 0.01 as the color difference eligibility standard.Experiments show that in the color calibration part,the pass rate of calibration using the initial calibration algorithm is 91.25%,and the pass rate of calibration using the Cor-Calibration algorithm is 99.79%,which verifies the superiority of the Cor-Calibration algorithm for high-precision calibration of RGB-LEDs.Comprehensive analysis of the color difference prediction algorithm and color difference calibration algorithm of RGB-LED,the intelligent prediction and correction method of RGB-LED chromaticity parameters has certain theoretical significance and good engineering application value in the field of color difference prediction and calibration.