LED light source uniformity and color deviation detection method

Abstract

The invention provides an LED light source uniformity and color deviation detection method, which comprises the following steps: setting environmental parameters of a constant temperature and humidity test box, monitoring the environment in the box in real time through a temperature and humidity sensor, and triggering a temperature and humidity control system to adjust when detecting that the temperature and humidity deviate from a set value; according to the initial color characteristic database, placing an LED light source in the constant-temperature and constant-humidity test box for an accelerated aging test, and obtaining spectral data of the light source at different time points by adopting a spectrum analyzer; calculating chromaticity coordinates, color temperatures and color rendering indexes corresponding to the spectral data at different time points, calculating a color deviation value of the LED light source, and fitting a change curve of the color deviation along with time; and according to an accelerated aging test result of the LED light source, calculating the color maintenance life of the light source in a high-temperature and high-humidity environment, and completing color deviation detection through a color stability uniformity index of the LED light source.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for detecting uniformity and color deviation of an LED light source. Background Art

[0002] In some indoor environments with high temperature and humidity, such as kitchens and bathrooms, LED lamps may accelerate the aging of packaging materials due to high humidity and high temperature, thus affecting their performance and life. In the manufacturing process of LED light sources, the impact of high temperature and high humidity on color stability is a technical problem that needs to be solved urgently. As the temperature and humidity increase, the quantum well structure inside the LED chip will undergo microscopic changes, causing the emission wavelength to drift. This color deviation not only exists between different saturations, but also changes over time, posing a challenge to product quality control. Traditional accelerated aging test methods are difficult to fully simulate the complex factors in the actual use environment, such as temperature and humidity cycle changes, light intensity fluctuations, etc. In addition, there is currently a lack of unified standards and measurement methods for accurately quantifying the color deviation of LED light sources, and data from different batches and manufacturers are difficult to compare directly. In high temperature and high humidity environments, there are differences in the aging rates of LED chips and packaging materials. This asynchronous aging further exacerbates the unpredictability of color deviation. How to accurately simulate the complex factors of the actual use environment in accelerated aging tests, establish a scientific and reasonable color deviation evaluation system, and optimize the material selection and structural design of LED light sources on this basis to improve their color stability in high temperature and high humidity environments is an important technical challenge facing the current LED industry. Summary of the invention

[0003] The present invention provides a method for detecting uniformity and color deviation of an LED light source, which mainly includes:

[0004] Set the environmental parameters of the constant temperature and humidity test chamber, and use the temperature and humidity sensors to monitor the environment in the chamber in real time. When the temperature and humidity are detected to deviate from the set values, the temperature and humidity control system is triggered to adjust;

[0005] Perform initial spectrum measurement on LED light sources, obtain the chromaticity coordinates, color temperature, and color rendering index of the light source, convert the measurement data into numerical values ​​in the standard color space, and establish an initial color characteristic database for LED light sources;

[0006] According to the initial color characteristic database, the LED light source is placed in the constant temperature and humidity test chamber for accelerated aging test, and the spectrum data of the light source at different time points is obtained using a spectrum analyzer;

[0007] Calculate the chromaticity coordinates, color temperature, and color rendering index corresponding to the spectral data at different time points, calculate the color deviation of the LED light source, and fit the curve of color deviation changing with time;

[0008] Perform dimensionality reduction processing on the color deviation data of LED light sources, extract factors that affect color stability, establish a color compensation model, input the working time and environmental parameters of the LED light source, output color compensation parameters, and dynamically compensate the color of the LED light source;

[0009] The current and voltage parameters of the light source are adjusted through the LED driving circuit, the color of the LED light source is dynamically compensated, the intensity ratio of light of different wavelengths is adjusted, the color rendering index and color saturation of the LED light source are improved, the color output of the LED light source is monitored in real time, and when the color saturation parameter is detected to deviate from the target value, corresponding correction is performed;

[0010] According to the accelerated aging test results of LED light sources, the color maintenance life of the light source in a high temperature and high humidity environment is calculated, and the color deviation detection is completed through the color stability uniformity index of the LED light source.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The invention discloses a method for detecting uniformity and color deviation of an LED light source. The method simulates different environmental conditions through a constant temperature and humidity test chamber, performs accelerated aging tests on LED light sources, and monitors the spectrum changes in real time. Color analysis software is used to calculate chromaticity parameters, fit the curve of color deviation over time, extract the main factors affecting color stability, and establish a color compensation model. The invention dynamically outputs color compensation parameters according to LED working time and environmental parameters, realizes color compensation by adjusting driving circuit parameters, and improves color rendering index and color saturation. At the same time, multiple small test chambers are used for parallel testing to expand the test capacity. The invention realizes accurate testing and dynamic compensation of color stability of LED light sources, effectively prolongs the color maintenance life of LED light sources, improves color stability and uniformity, and provides technical support for LED product quality control. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a method for detecting uniformity and color deviation of an LED light source.

[0014] Figure 2 The figure is a schematic diagram of a method for detecting uniformity and color deviation of an LED light source according to the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0016] like Figure 1 -2. In this embodiment, a method for detecting uniformity and color deviation of an LED light source may specifically include:

[0017] Step S101, setting the environmental parameters of the constant temperature and humidity test box, monitoring the environment in the box in real time through the temperature and humidity sensor, and triggering the temperature and humidity control system to adjust when it is detected that the temperature and humidity deviate from the set value.

[0018] Acquire the setting value of the environmental parameter, which is read from a pre-established database; collect the real-time temperature and humidity data in the test box, which is collected by a temperature and humidity sensor array at multiple points; perform sliding average filtering on the real-time temperature and humidity data to obtain the measured value of temperature and humidity. Calculate the deviation between the measured value of temperature and humidity and the set value of temperature and humidity; determine whether the deviation exceeds a preset threshold, and if the deviation exceeds the preset threshold, trigger parameter control. According to the size of the deviation value and the change trend, calculate the heater power adjustment amount and the humidifier water adjustment amount, and the heater power adjustment amount and the humidifier water adjustment amount are output by the PID controller.

[0019] Exemplarily, according to the preset temperature and humidity range of the constant temperature and humidity test chamber, the corresponding environmental parameter setting values ​​are obtained from the database, and the temperature setting values ​​and humidity setting values ​​are input into the temperature and humidity control system through the control panel. The control system initializes the heater power and humidifier water volume according to the setting values. A high-precision temperature and humidity sensor array is used to collect real-time temperature and humidity data at multiple points in the test chamber. The sensor acquisition frequency is set to once per second. The collected temperature and humidity values ​​are subjected to sliding average filtering, and the average value within 5 seconds is taken as the current temperature and humidity measured value. The deviation between the measured temperature and humidity values ​​and the set value is calculated by a comparison algorithm. If the temperature deviation exceeds ±0.05℃ or the humidity deviation exceeds ±0.5%RH, the incremental PID controller is triggered. The proportional coefficient Kp of the PID controller is set to 0.5, the integral coefficient Ki is set to 0.1, and the differential coefficient Kd is set to 0.05. The controller calculates the heater power adjustment amount and the humidifier water adjustment amount according to the deviation value and the change trend. The temperature and humidity control actuator receives the adjustment amount output by the PID controller. For the heater, PWM modulation is used to control the power output. The PWM cycle is set to 1 second and the duty cycle accuracy is 0.1%. For the humidifier, a stepper motor is used to control the speed of the water pump. The minimum step angle of the stepper motor is 1.8°. 128 subdivision control is achieved through micro-step drive. The response time of the control actuator is set to 1 second, and the adjustment accuracy can reach ±0.05℃ for temperature and ±0.5%RH for humidity. The control system of the constant temperature and humidity test chamber uses a relational database based on MySQL to store the preset temperature and humidity range parameters. The control panel communicates with the database server through the TCP / IP protocol to obtain the temperature setting value of 25℃ and the humidity setting value of 60%RH required by the current test. The control system uses a 32-bit ARM processor and runs the real-time operating system FreeRTOS. It outputs a 0-5V analog signal to initialize the heater power to 50% and the humidifier water volume to 30%. Eight SHT35 digital temperature and humidity sensors are arranged in the test chamber, and the I2C bus is used for data transmission, and data is collected once per second. The temperature and humidity values ​​are represented by 32-bit floating point numbers, and the data of the last 5 seconds are subjected to sliding average filtering. The filtered temperature value is 25.03℃ and the humidity value is 59.8%RH. The comparison algorithm uses a simple subtraction operation to calculate the temperature deviation of 0.03℃ and the humidity deviation of -0.2%RH. Since the temperature deviation does not exceed the ±0.05℃ threshold and the humidity deviation does not exceed the ±0.5%RH threshold, the PID controller is not triggered at this time. If the temperature rises to 25.06℃ and exceeds the threshold in subsequent monitoring, the incremental PID controller is triggered. The PID controller uses floating point operations and calculates that the heater power needs to be reduced by 0.3% based on the set proportional coefficient of 0.5, integral coefficient of 0.1 and differential coefficient of 0.05.The heater in the temperature and humidity control actuator uses MOSFET as the power output element. The PWM signal is generated by the timer. The clock frequency is 10MHz, the PWM cycle is 1 second, and it corresponds to 10,000,000 clock cycles. The duty cycle accuracy is 0.1%, which is equivalent to 10,000 discrete levels. The humidifier water pump is driven by a 42 stepper motor, using the A4988 stepper motor driver chip. By setting the MS1, MS2, and MS3 pins, 128 subdivision control is achieved, and the minimum adjustment angle is 0.014°. The response time of the control actuator is 1 second, during which the PWM duty cycle adjustment and the stepper motor speed change are completed to achieve precise adjustment of temperature and humidity.

[0020] Step S102 , performing an initial spectrum measurement on the LED light source, obtaining the chromaticity coordinates, color temperature, and color rendering index of the light source, converting the measurement data into numerical values ​​in a standard color space, and establishing an initial color characteristic database of the LED light source.

[0021] Acquire the spectrum data of the LED light source, which is collected by a high-precision spectrometer within a preset wavelength range. Calculate the tristimulus values ​​X, Y, and Z based on the spectrum data; determine the correlated color temperature using the MacAdam method; select several standard color samples and calculate the color rendering index Ra, which is the average of the color rendering indexes of several standard color samples. Convert the chromaticity coordinates into coordinates in a uniform color space. Create an initial color characteristic data table for the LED light source, which contains the fields of LED number, spectrum data, chromaticity coordinates, uniform color space coordinates, correlated color temperature, and color rendering index Ra. Establish a unique index for the LED number to improve the query efficiency of the data table.

[0022] Exemplarily, a high-precision spectrometer is used to measure the spectrum of an LED light source. The wavelength range of the spectrometer is 380-780nm, the wavelength resolution is 1nm, and the spectrum acquisition time is set to 100ms. Each LED light source is measured 10 times continuously, and the Simpson integration method is used to calculate the average spectrum data of the 10 measurements. The integration interval is 1nm to obtain the spectrum data of the LED light source. According to the acquired spectrum data, the three stimulus values ​​X, Y, and Z are calculated using the CIE1931 colorimetry standard observer function, and then the chromaticity coordinates (x, y) of the LED light source are calculated, where x = X / (X+Y+Z), y = Y / (X+Y+Z). The color temperature is calculated using the MacAdam method, and the closest point on the Planckian locus is found through iterative calculation. The color rendering index Ra is calculated using the CIE1995 method, 15 standard color samples are selected, the color rendering index of each color sample is calculated, and the average value is taken. The measured chromaticity coordinates (x, y) are converted to (u', v') coordinates of CIE1976 uniform color space, and the conversion formula is u' = 4x / (-2x+12y+3), v' = 9y / (-2x+12y+3). At the same time, the L*, a*, and b values ​​of CIELAB color space are calculated, and the conversion formula is L = 116f(Y / Yn)-16, a* = 500[f(X / Xn)-f(Y / Yn)], b* = 200[f(Y / Yn)-f(Z / Zn)], where Xn, Yn, and Zn are the three stimulus values ​​of the standard white point. The color temperature and color rendering index Ra values ​​are retained to two decimal places to form a value set of the standard color space. The MySQL relational database is used to create the initial color characteristic data table of LED light sources. The table structure includes the primary key ID (integer, auto-increment), LED number (character, 20 bits), spectrum data (BLOB type), chromaticity coordinates x, y (double precision floating point), u', v' coordinates (double precision floating point), correlated color temperature (integer), color rendering index Ra (double precision floating point), L*, a*, b* values ​​(double precision floating point) and other fields. The converted standard color space value set is written into the data table using a preprocessing statement, and a unique index of the LED number is established to improve the query efficiency, forming the initial color characteristic database of the LED light source. In the process of measuring the LED light source, the spectrum measurement is performed using the spectral radiometer OL770, the wavelength range is set to 380-780nm, the wavelength resolution is 1nm, and the integration time is set to 100ms. Each LED light source is measured 10 times continuously, and the spectrum data of 401 wavelength points are obtained each time. The Simpson integration method is used to average the 10 measurement results to obtain the final spectrum data. Taking a typical white light LED as an example, its main wavelength is 455nm, and the peak value of the phosphor is around 565nm. Using the CIE1931 standard chromaticity observer function, the spectrum data is integrated and calculated to obtain the three stimulus values ​​X=3256, Y=3412, and Z=1576.Then the chromaticity coordinates x=0.3952, y=0.4141 are calculated. The MacAdam method is used to calculate the correlated color temperature. After 5 iterations, the correlated color temperature of the LED is 3985K. 15 standard color samples are selected for color rendering index calculation, and the average color rendering index Ra is calculated to be 82. The chromaticity coordinates are converted to CIE1976UCS chromaticity diagram coordinates, and u'=0.2247, v'=0.5298 are obtained. At the same time, the CIELAB color space values ​​are calculated, and L*=65.32, a*=-12.45, b*=14.78 are obtained. All calculation results are rounded to two decimal places. Use MySQL database to store measurement results, create a table named led_color_characteristics, including id (primary key, auto-increment), led_code (varchar(20)), spectrum_data (blob), x (double), y (double), u_prime (double), v_prime (double), cct (int), cri (double), l_star (double), a_star (double), b_star (double), etc. Use prepared statements to insert data into the table, and create a unique index on the led_code field to improve the efficiency of subsequent queries.

[0023] Step S103, placing the LED light source in the constant temperature and humidity test box to perform an accelerated aging test according to the initial color characteristic database, and using a spectrum analyzer to obtain spectrum data of the light source at different time points.

[0024] The LED light source number list stored in the initial color characteristic database is obtained, and the LED light source is placed at the preset coordinates in the constant temperature and humidity test box according to the LED light source number list; an accelerated aging test program is executed, and the accelerated aging test program sets the total test time and the sampling time interval; the spectrum analyzer is controlled, and if the sampling time point is reached, the spectrum measurement is triggered, and the electrical parameters of the LED light source are monitored at the same time; the spectrum analyzer performs multiple measurements on each LED light source according to the preset measurement parameters, and uses a moving average filtering algorithm to process the multiple measurement data; the measured spectrum data, the accumulated aging time, and the environmental parameter information are associated with the LED light source number and inserted into a database table, and the database table contains the LED number, the measurement time point, the spectrum data, the temperature, the humidity, and the light source status fields.

[0025] Exemplarily, the LED light source number list is read from the initial color characteristic database, and the LED light source is placed on a fixed bracket in a constant temperature and humidity test chamber according to preset coordinates using an automated robotic arm. The bracket is equipped with an independently numbered interface to ensure that each LED light source is unique and traceable. The test chamber temperature is set to 85°C, and the relative humidity is set to 85%. The environmental parameters are monitored and recorded in real time through temperature and humidity sensors. Start the accelerated aging test program written in Python, set the total test time to 1000 hours, and the sampling time interval to 50 hours. The program controls the spectrum analyzer through serial communication, triggers spectrum measurement at each sampling time point, and the program monitors the electrical parameters of the LED light source in real time. If an abnormal current is detected, the LED is immediately recorded and marked as a fault state. The measurement parameters of the spectrum analyzer are set to a wavelength range of 380-780nm, a resolution of 1nm, and an integration time of 100ms. Each LED light source is measured continuously for 3 times, and the moving average filtering algorithm is used to process the data of the 3 measurements to remove possible noise interference, obtain stable spectrum data at that time point, and record the precise timestamp of the spectrum measurement, accurate to milliseconds, and record the current accumulated aging time. Use MySQL relational database to create aging test data table. The table structure includes primary key ID (integer, auto-increment), LED number (character, 20 bits), measurement time point (timestamp type), spectrum data (BLOB type), temperature (floating point type), humidity (floating point type), light source status (integer type) and other fields. Use preprocessing statements to associate the measured spectrum data, cumulative aging time, environmental parameters and other information with the LED light source number and insert them into the data table. A composite index is established on the LED number and the measurement time point to improve the efficiency of subsequent query and analysis. After each measurement at a time point is completed, the corresponding record in the database is updated. In the LED accelerated aging test, the initial color characteristic table is first read from the MySQL database to obtain the LED number list, such as LED001 to LED100. Use KUKAKR3R540 automated robotic arm with an accuracy of ±0.02mm to place the LEDs on 100 numbered interfaces in the constant temperature and humidity test chamber at preset coordinates (X: 100mm, Y: 100mm, Z: 50mm). The test chamber model is ESPECPSL-2KPH, the temperature is set to 85±0.3℃, the relative humidity is 85±1%RH, and the environmental data is collected every 10 seconds through the PT100 temperature sensor and the polymer humidity sensor. Start the test program written in Python, set the total duration to 1000 hours, and perform spectral measurements every 50 hours. The program communicates with the OceanOpticsQE65000 spectrum analyzer through the RS-232 serial port at a baud rate of 9600 to trigger the measurement. At the same time, the program monitors the LED current through the Keithley2400 source meter with a sampling rate of 1Hz. If the current fluctuation exceeds ±5%, the LED is marked as a fault state.The spectrum measurement parameters are set to a wavelength range of 380-780nm, a resolution of 1nm, and an integration time of 100ms. Each LED is measured three times in succession, and the data is processed using a 5-point moving average filter algorithm to obtain a smooth spectrum curve. The measurement is recorded with a timestamp accuracy of 1ms. The processed data is stored in the aging_test table of the MySQL database. The table structure includes id (INT), led_code (VARCHAR (20)), measure_time (TIMESTAMP), spectrum_data (BLOB), temperature (FLOAT), humidity (FLOAT), and led_status (TINYINT) fields. Use prepared statements to insert data, and create a composite index BTREE on the led_code and measure_time paths to optimize query speed.

[0026] Different LED light sources are tested in parallel using multiple constant temperature and humidity test chambers. According to the model and specifications of the LED light source, the standard spectrum data of the corresponding model of LED light source is obtained from the pre-established initial color characteristic data. Multiple different environmental parameter combinations are set in the constant temperature and humidity test chamber, and a test chamber of corresponding size is selected for testing.

[0027] Acquire standard spectral data from a pre-established initial color characteristic database, the standard spectral data corresponding to the model and specification of the LED light source; determine the spectral peak wavelength and color temperature of the LED light source according to the standard spectral data; classify the LED light sources using a clustering algorithm to obtain an LED light source group with similar spectral characteristics; generate an environmental parameter combination for the LED light source group, and set a different environmental parameter combination in each test box; acquire spatial position information of the LED light source; place the LED light source at a preset position in the test box; perform parallel testing of multiple test boxes, including synchronous aging testing of each test box; and receive environmental parameters and spectral data sent by a local data acquisition node.

[0028] Exemplarily, according to the model and specification of the LED light source, standard spectral data is extracted from the pre-established initial color characteristic database, and the LED light source is classified by the K-means clustering algorithm. The spectral peak wavelength and color temperature are used as characteristic values. The number of clusters is set to 5, the number of iterations is set to 100, and 5 groups of LED light sources with similar spectral characteristics are obtained. Each group is assigned a small constant temperature and humidity test box. For each group of LED light sources, the L9 (3 ^4) orthogonal table is used to design the environmental parameter combination. The factors include temperature 25 ℃, 55 ℃, 85 ℃ and humidity 45%, 65%, 85%, and 9 environmental parameter combinations are generated. Different environmental parameter combinations are set in each test box. A constant temperature and humidity test box with an internal size of 40cm×40cm×50cm is selected. 100 LED light sources are placed in each test box. A six-axis robotic arm is used to accurately grab the LED light source and place it in the preset fixture position in the test box. The fixture spacing is 2cm. Each test box is equipped with an independent temperature and humidity control unit and a spectrum measurement device to achieve parallel testing of multiple test boxes. The parallel test program was started, and each test box was synchronously tested for aging. A distributed data acquisition system based on the MQTT protocol was used. Each test box was equipped with a Raspberry Pi as a local data acquisition node. The data was transmitted to the central server through the MQTTBroker. The InfluxDB time series database was used to store environmental parameters and spectral data. Real-time data visualization was achieved through Grafana, and centralized management and analysis of multi-test box data was achieved. In the accelerated aging test of LED light sources, the standard spectral data of 1000 LED light sources were first extracted from the MySQL database and classified using the K-means clustering algorithm. The spectral peak wavelength, such as 450nm, 550nm and color temperature, such as 3000K, 6500K, was used as the eigenvalue, the number of cluster centers k=5, the maximum number of iterations was 100, and 5 groups of LED light sources were obtained. Each group was equipped with an ESPECSH-242 constant temperature and humidity test box with an internal size of 40cm×40cm×50cm. Nine combinations of environmental parameters were designed using the L9 (3^4) orthogonal table, with temperatures of 25°C, 55°C, and 85°C, and humidity of 45%, 65%, and 85%. The KUKAKR3R540 six-axis robotic arm, in conjunction with the Cognex In-Sight7000 vision system, accurately grasps the LED light source and places it in the test chamber, with a fixture spacing of 2 cm and 100 per box. The test chamber is equipped with a Vaisala HMT330 temperature and humidity sensor with an accuracy of ±0.1°C and ±1% RH, and an OceanOptics QE65000 spectrometer for spectral measurement. A distributed data acquisition system based on the MQTT protocol is used, with each test chamber using a RaspberryPi4B as a local node, and data is transmitted to the central server through EMQXBroker.InfluxDB is used to store time series data, with a retention policy set to 30 days, and Grafana creates a real-time dashboard to display temperature, humidity, and spectrum trends. Through this parallel testing method, the accelerated aging test of 5,000 LED samples was completed within 1,000 hours, greatly improving the test efficiency and comprehensiveness of data collection.

[0029] Step S104, calculating the chromaticity coordinates, color temperature, and color rendering index corresponding to the spectrum data at different time points, calculating the color deviation of the LED light source, and fitting a curve of the color deviation changing with time.

[0030] The spectral data of the LED light source at different time points are obtained, and the tristimulus values ​​X, Y, and Z corresponding to the spectral data are calculated to obtain the chromaticity coordinates (x, y); after the chromaticity coordinates (x, y) are determined, the correlated color temperature CCT and the color rendering index Ra are calculated; according to the chromaticity coordinates (x, y), the correlated color temperature CCT and the color rendering index Ra, the chromaticity coordinate deviation Δu'v', the color temperature deviation ΔCCT and the color rendering index deviation ΔRa are calculated; for the changes of Δu'v', ΔCCT and ΔRa over time, a third-order polynomial fitting is performed using the least squares method to obtain a fitting function; according to the fitting function, the corresponding color deviation value is calculated through a predefined time point.

[0031] Exemplarily, the spectrum data of LED light sources at different time points are extracted from the database, and the three stimulus values ​​X, Y, and Z are calculated using the CIE1931 colorimetry calculation formula to obtain the chromaticity coordinates (x, y). The correlated color temperature CCT is calculated using the MacAdam formula, and the color rendering index Ra is calculated using the CIE13.3-1995 method. 15 standard color samples are selected to calculate the special color rendering index Ri, and the average value is taken to obtain Ra. The calculation results are stored in the result data table. Based on the initial color characteristic data and the calculation results at each time point, the chromaticity coordinate deviation Δu'v' is calculated, and the CIE1976UCS chromaticity diagram is used for calculation. The color temperature deviation ΔCCT is calculated by

[0032] |CCT test - CCT initial| is obtained, and the color rendering index deviation ΔRa is calculated by |Ra test - Ra initial|. The deviation data is associated with the corresponding time point and stored in the color deviation data table. The polyfit function of the numpy library is used to perform least squares polynomial fitting. The changes of Δu'v', ΔCCT and ΔRa over time are respectively fitted by three polynomials to obtain a fitting function of the form at ^3+bt ^2+c*t+d, where t is time, a, b, c, and d are constants. The fitting parameters and goodness of fit R2 are stored in the fitting result table. Based on the fitting function, the poly1d function of numpy is used to create a polynomial object. The corresponding color deviation value is calculated by passing in predefined time points, such as 100 hours, 1000 hours, and 10000 hours. At the same time, the fsolve function of the scipy.optimize library is used to solve when the fitting function reaches the preset deviation threshold. The prediction results are saved in the prediction data table to provide data support for subsequent analysis and decision-making. In the process of analyzing the color performance of LED light sources, the spectral data of four time points, 0 hours, 100 hours, 500 hours and 1000 hours, are first extracted from the MySQL database. The three stimulus values ​​are calculated using the CIE1931 colorimetric formula, such as X=3256, Y=3412, Z=1576 at 0 hours, and the chromaticity coordinates x=0.3952, y=0.4141. The correlated color temperature CCT is calculated using the MacAdam formula, which is 3985K at 0 hours. 15 standard color samples are selected for the color rendering index Ra calculation, and Ra is 82 at 0 hours. The results are stored in the results data table. The color deviation is calculated, such as Δu'v'=0.0056, ΔCCT=127K, ΔRa=5 at 1000 hours, and stored in the deviations data table. Use the polyfit function of the numpy library to perform a cubic polynomial fit, such as Δu'v'(t) = 2e-11t'3-3e-8t'2+2e-5t+0.0001, R2 = 0.9985, e is a constant, and store it in the fitting_results data table. Based on the fitting function, predict the color deviation at 10,000 hours, such as Δu'v' = 0.0189, ΔCCT = 423K, ΔRa = 16. Use the scipy.optimize.fsolve function to solve the time point when Δu'v' reaches 0.01, and get 5783 hours. All prediction results are stored in the predictions prediction data table.

[0033] Step S105, performing dimensionality reduction processing on the color deviation data of the LED light source, extracting factors affecting color stability, establishing a color compensation model, inputting the working time and environmental parameters of the LED light source, outputting color compensation parameters, and dynamically compensating the color of the LED light source.

[0034] The color deviation data of the LED light source and the corresponding working time and environmental parameters are obtained, wherein the color deviation data includes chromaticity coordinate deviation, color temperature deviation and color rendering index deviation; the color deviation data is subjected to dimensionality reduction processing, and if the cumulative contribution rate of the eigenvalue and the eigenvector reaches a preset threshold, the main component is determined; a nonlinear color compensation model is constructed according to the main component, wherein the nonlinear color compensation model adopts a support vector regression algorithm and selects a radial basis function as a kernel function; the nonlinear color compensation model is converted into a lightweight format, and the nonlinear color compensation model in the lightweight format is downloaded to an LED driving controller; the working time and environmental parameters sent by the LED driving controller are received, the color compensation parameters are calculated according to the nonlinear color compensation model, and the duty cycle of the LED driving current of the three RGB channels is adjusted.

[0035] Exemplarily, the color deviation data of LED light sources are extracted from the database, including chromaticity coordinate deviation, color temperature deviation, color rendering index deviation, and corresponding working time and environmental parameters. The principal component analysis algorithm is used to reduce the dimension of the data. The PCA method is used to achieve this. It is set to retain 95% of the variance information. By calculating the eigenvalues ​​and eigenvectors, the main components are selected to extract the main factors affecting color stability. Based on the reduced-dimensional data, a nonlinear color compensation model is constructed using the support vector regression algorithm. The SVR algorithm is used to select the RBF kernel function. The hyperparameters C and gamma are optimized through grid search. The working time and environmental parameters are used as input variables, and the color compensation parameters are used as output variables. The constructed color compensation model is converted into a lightweight format, and the model parameters are downloaded to the ARMCortex-M4-based LED driver controller through serial communication. The controller uses a real-time operating system, samples the working time and environmental parameters every 100ms, receives the working time and environmental parameter information of the LED light source in real time, and inputs the model to calculate the color compensation parameters. According to the calculated color compensation parameters, the PWM technology is used to adjust the LED driving current of the three channels of RGB. The PWM frequency is set to 20kHz and the resolution is 12 bits. By adjusting the duty cycle, 0.025% fine adjustment is achieved to achieve dynamic compensation of the color of the LED light source and maintain the stability of the color output. In the LED color compensation system, 1000 sets of color deviation data of LED light sources are first extracted from the MySQL database, including Δu'v', ΔCCT, ΔRa, and the corresponding working time 0-10000 hours and ambient temperature -20℃ to 60℃. The PCA class of the sklearn library is used for principal component analysis, and n_components=0.95 is set to retain 95% of the variance information. After PCA processing, the original 5-dimensional data is reduced to 3 dimensions, and the main influencing factors are working time, ambient temperature and Δu'v'. Based on the reduced-dimensional data, the SVR class of the sklearn library was used to build a nonlinear color compensation model, the RBF kernel function was selected, and the parameters were optimized through GridSearchCV, and finally C=10 and gamma=0.1 were determined. After the model training was completed, TensorFlowLite was used to convert the model into an 8-bit quantized .tflite format, and the size was compressed from the original 5MB to 500KB. The model was downloaded to the LED driver controller based on STM32F407 through the UART interface at a baud rate of 115200. The controller runs the FreeRTOS environment, samples the ambient temperature through the ADC analog-to-digital converter every 100ms, and reads the working time from the RTC real-time clock. The sampled data is input into the nonlinear color compensation model, and the compensation parameters of the RGB three channels are calculated.The controller uses the Timer1, Timer2, and Timer3 clocks to generate a 20kHz, 12-bit resolution PWM signal, and dynamically adjusts the duty cycle according to the compensation parameters to achieve a fine adjustment of 0.025%. In this way, the color deviation of the LED light source can be compensated in real time, and Δu'v' can be controlled within 0.003 within a 10,000-hour service life, ensuring the long-term color stability of the LED light source.

[0036] Step S106, adjusting the current and voltage parameters of the light source through the LED driving circuit, dynamically compensating the color of the LED light source, adjusting the intensity ratio of light of different wavelengths, and improving the color rendering index and color saturation of the LED light source; real-time monitoring of the color output of the LED light source, and performing corresponding corrections when it is detected that the color saturation parameter deviates from the target value.

[0037] Obtain the resistance range and adjustment level of a digital potentiometer, the digital potentiometer being used to control the output current of a driver chip; control the brightness of an LED according to the output current of the driver chip, the brightness of the LED being adjusted by PWM modulation; use a spectrum analyzer to collect spectrum data of the LED light source, the spectrum data being used to calculate a color rendering index and color saturation; run a PID control algorithm, the PID algorithm using a target color rendering index and color saturation as set values ​​and an actual measured value as a feedback input; determine whether an output compensation value of the PID algorithm is within a preset range; if the compensation value is within the preset range, map the compensation value to the resistance value of the digital potentiometer and the PWM duty cycle; adjust the resistance value and the PWM duty cycle of the digital potentiometer, the adjustment being used to change the current and brightness of each LED channel.

[0038] For example, a multi-channel LED drive circuit is designed, using an MCP4151 digital potentiometer with a resistance range of 0-10kΩ and 256 levels of adjustment. The LED current is controlled by an LM3405 driver chip with a range of 0-1A. The brightness of the LED is controlled by PWM modulation, and the PWM frequency is set to 20kHz with a resolution of 12 bits to achieve a fine adjustment of 0.025%. The spectrum data of the LED light source is collected in real time using an OceanOpticsQE65000 spectrum analyzer with a measurement range of 350-1000nm and a resolution of 0.8nm. The color rendering index is calculated using the CIE13.3-1995 method, and the color saturation is calculated using the CIECAM02 color appearance model. The sampling frequency is set to 10Hz, and the data is transmitted to the control unit via a USB interface. The incremental PID control algorithm is run in the ArduinoMega2560 control unit with an operating frequency of 16MHz, Kp=0.5, Ki=0.2, Kd=0.1, and the target color rendering index and color saturation are set as the set values. The measured values ​​are used as feedback inputs to calculate the compensation value of each LED channel. The PID parameters are determined by the Ziegler-Nichols tuning method. According to the compensation value output by the PID algorithm, the output range of -100 to 100 is linearly mapped to the digital potentiometer resistance of 0-10kΩ and the PWM duty cycle of 0-100%. The analogWrite() function of Arduino is used to generate a PWM signal. The resistance value and PWM duty cycle of the MCP4151 digital potentiometer are adjusted in real time through the SPI interface, thereby changing the current and brightness of each LED channel, realizing dynamic compensation of the color of the LED light source and improving the color rendering index and color saturation. In the dynamic compensation of LED color, a four-channel LED driver circuit is first designed to control red, green, blue and white LEDs respectively. Each channel uses MCP4151 digital potentiometer and LM3405 driver chip, with current range of 0-1A, PWM frequency of 20kHz, and resolution of 12 bits. For example, the red LED is initially set to 700mA and 50% duty cycle. OceanOpticsQE65000 spectrum analyzer is used to collect LED spectrum in real time, with a measurement range of 350-1000nm and sampling every 100ms. CIE13.3-1995 method is used to calculate color rendering index Ra, with an initial value of 82; CIECAM02 model is used to calculate color saturation s, with an initial value of 1.2. ArduinoMega2560 is used as the control unit to run the incremental PID algorithm, Kp=0.5, Ki=0.2, Kd=0.1, target Ra is set to 90, and s is set to 1.5. When the actual Ra=85 and s=1.3 are detected, the PID output compensation value is -20. The -20 is linearly mapped to a 1.6kΩ increase in the digital potentiometer resistance, the initial 5kΩ becomes 6.6kΩ, and the PWM duty cycle is reduced by 10%, from 50% to 40%.The MCP4151 is updated at a rate of 1MHz via the SPI interface, and the PWM is adjusted via the analogWrite() function. This causes the red LED current to drop to about 530mA, and the brightness to decrease. At the same time, other channels are adjusted accordingly, such as the blue LED current is increased to 850mA and the duty cycle is increased to 60%. After multiple iterations of adjustment, Ra=89.5, s=1.48 are finally achieved, which are close to the target values, completing the dynamic color compensation of the LED light source.

[0039] Step S107, according to the accelerated aging test result of the LED light source, the color maintenance life of the light source in a high temperature and high humidity environment is calculated, and the color deviation detection is completed through the color stability uniformity index of the LED light source.

[0040] A constant temperature and humidity test chamber is used to set up an accelerated aging test environment for an LED light source, wherein the temperature of the test environment is a preset temperature value, the relative humidity of the test environment is a preset humidity value, and the test duration is a preset duration value; according to a preset time interval, a spectrometer is used to collect spectral data of the LED light source, wherein the spectral data includes luminous flux, chromaticity coordinates, correlated color temperature, and color rendering index; the color difference values ​​of the LED light source at different time points are calculated according to the spectral data, and if the color difference value exceeds a preset threshold, it is determined to be the end of the color maintenance life of the LED light source; an image of the light-emitting surface of the LED light source is obtained, and a chromaticity uniformity index is calculated according to the light-emitting surface image, and the chromaticity uniformity index is used to characterize the color stability uniformity of the LED light source; the maximum color difference, average color difference, and color difference standard deviation of the LED light source during the entire test process are calculated according to the spectral data, and the maximum color difference, the average color difference, and the color difference standard deviation are used to evaluate the color deviation of the LED light source.

[0041] Exemplarily, an accelerated aging test environment for LED light sources is set in a constant temperature and humidity test chamber, with a temperature of 85°C, a relative humidity of 85%, and a test duration of 1000 hours. The spectrum data of the LED light source is collected every 50 hours using an OceanOpticsQE65000 spectrometer, with a wavelength range of 380-780nm and a resolution of 0.8nm, and the luminous flux, chromaticity coordinates, correlated color temperature, and color rendering index are recorded. Based on the collected spectrum data, the color difference values ​​Δu'v' of the LED light source at different time points are calculated. When Δu'v' exceeds 0.007, it is determined to be the end of the color maintenance life of the LED light source. The linear interpolation method is used to calculate the accurate color maintenance life, and two adjacent time points where Δu'v' just exceeds 0.007 are found, and the accurate life time is obtained by linear interpolation. The Allied Vision Prosilica GT2000 high-resolution CCD camera was used to capture the LED light-emitting surface image, calculate the chromaticity uniformity index CUI, divide the image into 9 equal-area areas, calculate the average chromaticity coordinates of each area, and then calculate the maximum deviation of these 9 coordinates from the overall average coordinates to obtain the CUI value. The closer the CUI value is to 1, the better the color uniformity is. CUI is used as the color stability uniformity indicator of the LED light source. Based on the collected spectral data, the maximum color difference Δu'v'max, the average color difference Δu'v'avg and the color difference standard deviation σΔu'v' of the LED light source during the entire test process are calculated. The color deviation of the LED light source is evaluated by combining these three parameters. The color deviation detection report is generated using Python's pandas library, including time series data, statistical charts and summary data. The report is output in HTML format for easy viewing and sharing. In the accelerated aging test of the LED light source, the ESPECPL-3KPH constant temperature and humidity test chamber is set to 85℃ and 85%RH, and the test lasts for 1000 hours. The OceanOptics QE65000 spectrometer collects spectral data every 50 hours, with a wavelength range of 380-780nm and an integration time of 100ms. Initial measurements show that the LED luminous flux is 1000lm, chromaticity coordinates (0.3127, 0.3290), CCT 6500K, and CRI 80. At 850 hours, the measured Δu'v' is 0.0068, and at 900 hours it is 0.0074. Through linear interpolation, the color maintenance life is calculated to be 875 hours. The AlliedVision Prosilica GT2000 camera captures a 16M pixel image of the LED light-emitting surface with an exposure time of 10ms. The image processing algorithm divides it into a 3x3 grid and calculates the average (u', v') of each grid. The central area is (0.2103, 0.4997), the edge area is (0.2116, 0.5012), the maximum deviation is 0.0022, and the CUI is 0.9978.Using the pandas library to process 21 sets of measurement data within 1000 hours, we calculated Δu'v'max = 0.0074, Δu'v'avg = 0.0042, σΔu'v' = 0.0015. The generated HTML report contains time series charts, such as 0-1000 hours Δu'v'.

[0042] The report is published via a web server to facilitate remote access and team sharing of analysis results.

[0043] The above only lists some preferred embodiments of the present invention, but the present invention is not limited thereto, and many improvements and changes can be made. As long as the improvements and changes are made on the basis of the basic principles of the present invention, they should be regarded as falling within the protection scope of the present invention.

Claims

1. A method for detecting uniformity and color deviation of LED light sources, characterized in that: The method comprises: Set the environmental parameters of the constant temperature and humidity test chamber, and use the temperature and humidity sensors to monitor the environment in the chamber in real time. When the temperature and humidity are detected to deviate from the set values, the temperature and humidity control system is triggered to adjust; Perform initial spectrum measurement on LED light sources, obtain the chromaticity coordinates, color temperature, and color rendering index of the light source, convert the measurement data into numerical values ​​in the standard color space, and establish an initial color characteristic database for LED light sources; According to the initial color characteristic database, the LED light source is placed in the constant temperature and humidity test chamber for accelerated aging test, and the spectrum data of the light source at different time points is obtained using a spectrum analyzer; Calculate the chromaticity coordinates, color temperature, and color rendering index corresponding to the spectral data at different time points, calculate the color deviation of the LED light source, and fit the curve of color deviation changing with time; Perform dimensionality reduction processing on the color deviation data of LED light sources, extract factors that affect color stability, establish a color compensation model, input the working time and environmental parameters of the LED light source, output color compensation parameters, and dynamically compensate the color of the LED light source; The current and voltage parameters of the light source are adjusted through the LED driving circuit, the color of the LED light source is dynamically compensated, the intensity ratio of light of different wavelengths is adjusted, the color rendering index and color saturation of the LED light source are improved, the color output of the LED light source is monitored in real time, and when the color saturation parameter is detected to deviate from the target value, corresponding correction is performed; According to the accelerated aging test results of LED light sources, the color maintenance life of the light source in a high temperature and high humidity environment is calculated, and the color deviation detection is completed through the color stability uniformity index of the LED light source.

2. The method according to claim 1, wherein: The environmental parameters of the constant temperature and humidity test box are set, and the environment in the box is monitored in real time by a temperature and humidity sensor. When the temperature and humidity are detected to deviate from the set values, the temperature and humidity control system is triggered to adjust, including: Obtaining environmental parameter setting values, wherein the environmental parameter setting values ​​are read from a pre-established database; Collecting real-time temperature and humidity data in the test box, wherein the real-time temperature and humidity data is collected at multiple points by a temperature and humidity sensor array; Perform sliding average filtering on the real-time temperature and humidity data to obtain actual temperature and humidity values; Calculating the deviation between the measured temperature and humidity values ​​and the set temperature and humidity values; Determine whether the deviation exceeds a preset threshold, and if the deviation exceeds the preset threshold, trigger parameter control; According to the size and change trend of the deviation value, the heater power adjustment amount and the humidifier water volume adjustment amount are calculated, and the heater power adjustment amount and the humidifier water volume adjustment amount are output by the PID controller.

3. The method according to claim 1, wherein: The initial spectrum measurement of the LED light source is performed to obtain the chromaticity coordinates, color temperature, and color rendering index of the light source, and the measurement data is converted into a value in a standard color space to establish an initial color characteristic database of the LED light source, including: Acquire spectral data of the LED light source, wherein the spectral data is collected by a high-precision spectrometer within a preset wavelength range; Calculate tristimulus values ​​X, Y and Z based on the spectral data; The MacAdam method was used to determine the correlated color temperature; Select several standard color samples and calculate the color rendering index Ra, where the color rendering index Ra is the average value of the color rendering indexes of the several standard color samples; Convert chromaticity coordinates to coordinates in a uniform color space; Create an initial color characteristic data table of the LED light source, the data table including fields of LED number, spectrum data, chromaticity coordinates, uniform color space coordinates, correlated color temperature and color rendering index Ra; A unique index of the LED number is established to improve the query efficiency of the data table.

4. The method according to claim 1, wherein: The method of placing the LED light source in the constant temperature and humidity test chamber for accelerated aging test according to the initial color characteristic database, and using a spectrum analyzer to obtain spectrum data of the light source at different time points includes: Obtaining a list of LED light source numbers stored in an initial color characteristic database, and placing the LED light sources at preset coordinates in a constant temperature and humidity test chamber according to the list of LED light source numbers; Executing an accelerated aging test program, wherein the accelerated aging test program sets a total test duration and a sampling time interval; Control the spectrum analyzer. If the sampling time point is reached, the spectrum measurement is triggered and the electrical parameters of the LED light source are monitored at the same time. The spectrum analyzer performs multiple measurements on each LED light source according to preset measurement parameters, and uses a moving average filtering algorithm to process multiple measurement data; Insert the measured spectral data, cumulative aging time and environmental parameter information into a database table in association with the LED light source number, wherein the database table includes fields for LED number, measurement time point, spectral data, temperature, humidity and light source status; It also includes: using multiple constant temperature and humidity test chambers for parallel testing to test different LED light sources, obtaining standard spectrum data of the corresponding model of LED light source from pre-established initial color characteristic data according to the model and specification of the LED light source, setting multiple different environmental parameter combinations in the constant temperature and humidity test chamber, and selecting a test chamber of corresponding size for testing.

5. The method according to claim 4, wherein: The method of testing different LED light sources by parallel testing in multiple constant temperature and humidity test chambers is described. According to the model and specification of the LED light source, the standard spectrum data of the corresponding model of LED light source is obtained from the pre-established initial color characteristic data. A plurality of different environmental parameter combinations are set in the constant temperature and humidity test chamber, and a test chamber of corresponding size is selected for testing, including: Acquire standard spectrum data from a pre-established initial color characteristic database, wherein the standard spectrum data corresponds to the model and specification of the LED light source; Determine the spectral peak wavelength and color temperature of the LED light source according to the standard spectral data; A clustering algorithm is used to classify LED light sources to obtain LED light source groups with similar spectral characteristics; Generating an environmental parameter combination for the LED light source group, wherein a different environmental parameter combination is set in each test box; Obtain the spatial position information of the LED light source; Placing the LED light source at a preset position in the test box; Conduct parallel testing of multiple chambers, including simultaneous aging testing of each chamber; Receive environmental parameters and spectral data sent by local data acquisition nodes.

6. The method according to claim 1, wherein: The method of calculating the chromaticity coordinates, color temperature, and color rendering index corresponding to the spectrum data at different time points, calculating the color deviation of the LED light source, and fitting the curve of the color deviation changing with time includes: Obtaining spectral data of LED light sources at different time points, calculating tristimulus values ​​X, Y and Z corresponding to the spectral data, and obtaining chromaticity coordinates (x, y); After the chromaticity coordinates (x, y) have been determined, the correlated color temperature CCT and the color rendering index Ra are calculated; Calculate the chromaticity coordinate deviation Δu'v', the color temperature deviation ΔCCT and the color rendering index deviation ΔRa according to the chromaticity coordinates (x, y), the correlated color temperature CCT and the color rendering index Ra; For the changes of Δu'v', ΔCCT and ΔRa over time, a third-order polynomial fitting is performed using the least squares method to obtain a fitting function; According to the fitting function, the corresponding color deviation value is calculated through the predefined time point.

7. The method according to claim 1, wherein: The color deviation data of the LED light source is subjected to dimensionality reduction processing, factors affecting color stability are extracted, a color compensation model is established, working time and environmental parameters of the LED light source are input, color compensation parameters are output, and the color of the LED light source is dynamically compensated, including: Acquire color deviation data of the LED light source and corresponding working time and environmental parameters, wherein the color deviation data includes chromaticity coordinate deviation, color temperature deviation and color rendering index deviation; Performing dimensionality reduction processing on the color deviation data, and determining the main components if the cumulative contribution rate of the eigenvalues ​​and eigenvectors reaches a preset threshold; Constructing a nonlinear color compensation model according to the main components, wherein the nonlinear color compensation model adopts a support vector regression algorithm and selects a radial basis function as a kernel function; Converting the nonlinear color compensation model into a lightweight format, and downloading the nonlinear color compensation model in the lightweight format to an LED drive controller; The working time and environmental parameters sent by the LED driving controller are received, the color compensation parameters are calculated according to the nonlinear color compensation model, and the duty ratios of the LED driving currents of the three RGB channels are adjusted.

8. The method according to claim 1, wherein: The LED driving circuit adjusts the current and voltage parameters of the light source, dynamically compensates for the color of the LED light source, adjusts the intensity ratio of light of different wavelengths, improves the color rendering index and color saturation of the LED light source, monitors the color output of the LED light source in real time, and performs corresponding correction when it is detected that the color saturation parameter deviates from the target value, including: Obtaining the resistance range and adjustment level of a digital potentiometer, wherein the digital potentiometer is used to control the output current of a driver chip; The brightness of the LED is controlled according to the output current of the driving chip, and the brightness of the LED is adjusted by PWM modulation; A spectrum analyzer is used to collect spectrum data of the LED light source, and the spectrum data is used to calculate the color rendering index and color saturation; Running a PID control algorithm, wherein the PID algorithm uses a target color rendering index and color saturation as set values ​​and uses a measured value as a feedback input; Determine whether the output compensation value of the PID algorithm is within a preset range; If the compensation value is within a preset range, mapping the compensation value to a digital potentiometer resistance value and a PWM duty cycle; The resistance value and PWM duty cycle of the digital potentiometer are adjusted, and the adjustment is used to change the current and brightness of each LED channel.

9. The method according to claim 1, wherein: The method calculates the color maintenance life of the light source in a high temperature and high humidity environment according to the accelerated aging test results of the LED light source, and completes the color deviation detection through the color stability uniformity index of the LED light source, including: A constant temperature and humidity test chamber is used to set an accelerated aging test environment for an LED light source, wherein the temperature of the test environment is a preset temperature value, the relative humidity of the test environment is a preset humidity value, and the test duration is a preset duration value; According to a preset time interval, a spectrometer is used to collect spectral data of the LED light source, wherein the spectral data includes luminous flux, chromaticity coordinates, correlated color temperature and color rendering index; Calculate the color difference value of the LED light source at different time points according to the spectral data, and if the color difference value exceeds a preset threshold, determine that the color maintenance life of the LED light source has ended; Acquire a light-emitting surface image of an LED light source, and calculate a chromaticity uniformity index according to the light-emitting surface image, wherein the chromaticity uniformity index is used to characterize the color stability uniformity of the LED light source; The maximum color difference, average color difference and color difference standard deviation of the LED light source in the entire test process are calculated according to the spectral data, and the maximum color difference, the average color difference and the color difference standard deviation are used to evaluate the color deviation of the LED light source.

Images