Neural Network-based CAN Bus RGB-LED Color Calibration System Research

With the continuous development of intelligent cockpit technology and LED lighting technology,RGB-LED has gradually developed from a lighting tool to an intelligent interactive product,and the gradual increase in the number of RGB-LEDs used has led to strict requirements for their color consistency.Due to the influence of LED material,production process,junction temperature and control chip data precision,even RGB-LEDs of the same model are lit with the same PWM duty cycle,they may have a certain degree of color difference.The current RGB-LED light color detection and calibration technology has the problems of low detection efficiency and low calibration accuracy.In order to solve these shortcomings,this paper proposes a neural network-based RGB-LED light color calibration system for CAN bus.The main work results and innovations of this paper include three main parts.The first is a color parameter prediction model based on Light GBM.By using the idea of gradient boosting decision tree,the function of predicting whether an LED is qualified or not can be achieved by testing only a small number of parameters.By analyzing the relationship between the luminous characteristic parameters such as forward voltage,forward current,junction temperature,main wavelength and light intensity of RGB-LEDs,the main causes of color errors and calibration methods of RGB-LEDs are summarized,and the features used for model training are proposed.In this paper,the proposed model is compared with the four models of decision tree,random forest,XGBoost and RF-XGBoost in four aspects of False positive,Recall,Precision and F1＿score,and the false detection rate is as low as 10.46% and the recall rate reaches 96.33%,proving that The Light GBM-based color parameter prediction model is able to have fewer false positives and higher prediction accuracy.The second is the fast color parameter calibration algorithm,which integrates the influence of current on RGB-LED wavelength and calculates the theoretical PWM after linear compensation of chromaticity coordinates.In order to solve the shortcomings of traditional color parameter calibration methods in dealing with the problem that the chromaticity coordinates are outside the color gamut after calibration,such as difficulty in taking the optimal solution,difficulty in choosing the traversal step and long time consumption,this paper takes the CIE1931 chromaticity space as the The algorithm compares the number of intersection points between the circle with the chromaticity coordinate point as the center and the error tolerance as the radius and the color gamut boundary to determine whether the RGB-LED meets the calibration condition.The effective improvement in speed and efficiency as well as the advantage in accuracy of this algorithm is demonstrated through experiments.Finally,the design and construction of the CAN bus-based RGB-LED light color calibration system.In order to verify the effectiveness of the color parameter prediction model and high-precision parameter calibration algorithm proposed in this paper,an automated calibration platform is designed and built in this paper,using servo motors and mechanical axis conveyors to realize the automation function,and using USB-LIN adapters,color analyzers and probes to realize the RGB-LED LED control,calibration and parameter acquisition using USB-LIN adapter,color analyzer and probe,and the two-way traceability of experimental data.For the CAN bus load factor,a Hopfield neural network is used for signal distribution optimization analysis.