A method for adaptive brightness adjustment of LED backlight source

Abstract

The present application provides an adaptive brightness adjustment method for an LED backlight source, comprising: obtaining the intensity and spectral distribution information of downward natural light and indoor lamp light sources in real time through a light sensor, superimposing the light sources, calculating the total light intensity and spectral distribution, judging whether an adaptive adjustment condition is met, and triggering the LED backlight source brightness adjustment process if the condition is met; detecting the actual brightness of each unit to be adjusted in real time, comparing the actual brightness with an optimized target brightness value, calculating the brightness error, introducing a machine learning algorithm to learn and predict the error data in real time, and dynamically adjusting the PWM duty cycle of the unit to be adjusted; continuously obtaining user evaluation feedback data on display quality, performing cluster analysis on the evaluation feedback data through an optimized unsupervised learning algorithm, and generating a brightness adjustment curve that adapts to the preferences of different users.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for adaptively adjusting brightness of an LED backlight source. Background Art

[0002] In practical applications, the LED backlight adaptive brightness adjustment system faces complex light source superposition and spatial coupling problems. First, the downward natural light and the indoor lamp light source will produce a superposition effect of light intensity and spectral distribution, resulting in uneven light field distribution, which brings challenges to adaptive brightness adjustment. Secondly, the indoor lighting structure, such as the position and size of windows and skylights, as well as the layout position and angle of lamps, will form a complex coupling relationship with the indoor space, further affecting the light field distribution. Furthermore, indoor metal decorative components such as metal ceilings and stainless steel decorative panels will produce multiple reflections and modulation effects on light, change the propagation path and intensity distribution of light, and increase the difficulty of adaptive brightness adjustment. Finally, the composite light field formed by the combined action of the above factors will have dynamic changes in light intensity, spectral distribution, spatial distribution and other characteristics, which puts higher requirements on the perception, decision-making and execution of the adaptive brightness adjustment system. How to achieve accurate adaptive brightness adjustment of LED backlight sources in a complex light field environment is a technical problem that needs to be solved urgently. Summary of the invention

[0003] The present invention provides an adaptive brightness adjustment method for an LED backlight source, which mainly includes:

[0004] The intensity and spectral distribution information of downward natural light and indoor lamps are respectively obtained in real time through light sensors, and the light sources are superimposed to calculate the total light intensity and spectral distribution, and determine whether the adaptive adjustment conditions are met. If so, the LED backlight brightness adjustment process is triggered;

[0005] In the LED backlight brightness adjustment process, a spatial coupling model is established based on the indoor lighting structure, including the position and size of windows and skylights, as well as the layout position and angle of lamps, to analyze the factors affecting the diffuse distribution of natural light through the curtains, optimize the calculation method of diffuse light intensity distribution, and obtain the ambient light intensity modulation coefficient of different areas;

[0006] According to the spatial partition, the LED backlight source is divided into several independent units to be adjusted. According to the ambient light intensity modulation coefficient, the target brightness value of each unit to be adjusted is determined respectively, and the brightness of each unit to be adjusted is gradually adjusted using a smooth transition method to achieve the target brightness value;

[0007] Detect the actual brightness of each unit to be adjusted in real time, compare the actual brightness with the optimized target brightness value, calculate the brightness error, introduce machine learning algorithm to learn and predict the error data in real time, and dynamically adjust the PWM duty cycle of the unit to be adjusted;

[0008] Continuously obtain user evaluation feedback data on display quality, perform cluster analysis on the evaluation feedback data through an optimized unsupervised learning algorithm, and generate a brightness adjustment curve that adapts to different user preferences;

[0009] In view of the dynamic change characteristics of the composite light field, a prediction model of total light intensity, spectral distribution, and spatial distribution is established through machine learning algorithms to predict the change trend of the light field within a preset time period in advance to obtain the composite light field prediction results, and the prediction results are used to adaptively optimize the adjustment strategy of the backlight source;

[0010] Composite light field prediction is performed based on real-time collected multi-light sensor data. At the same time, the spatial brightness distribution of the current display screen is used as the optimization objective function. The PWM duty cycle of each backlight unit to be adjusted is solved in real time through the reinforcement learning algorithm to obtain the dynamically optimized backlight brightness adjustment strategy, which is transmitted to the LED backlight drive circuit to execute the corresponding brightness adjustment action.

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

[0012] The invention discloses an adaptive brightness adjustment method for an LED backlight source. The intensity and spectral distribution information of natural light and indoor light sources are obtained in real time by a light sensor, and a light source superposition model is established to calculate the total light intensity. According to the indoor lighting structure and lamp arrangement, a spatial coupling model is established to analyze the diffuse distribution of natural light, and the ambient light intensity modulation coefficient is obtained. The LED backlight source is divided into a plurality of independent adjustment units, the target brightness of each unit is determined according to the ambient light intensity modulation coefficient, and the brightness is adjusted by a smooth transition method. A machine learning algorithm is introduced to learn and predict the brightness error in real time, and the PWM duty cycle is dynamically adjusted. Cluster analysis is performed on user evaluation feedback by an unsupervised learning algorithm to generate an adjustment curve adapted to the preferences of different users. A composite light field prediction model is established to estimate the light field change trend in advance. Multi-sensor data and the current picture brightness distribution are used as input, and the backlight brightness adjustment strategy is optimized in real time by a reinforcement learning algorithm. The invention realizes the intelligent adaptive adjustment of the LED backlight source, and improves the display quality and user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a method for adaptively adjusting brightness of an LED backlight source.

[0014] Figure 2It is a schematic diagram of an adaptive brightness adjustment method of an LED backlight source according to the present invention.

[0015] Figure 3 It is another schematic diagram of an adaptive brightness adjustment method of an LED backlight source according to the present invention. DETAILED DESCRIPTION

[0016] 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 described clearly and completely 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.

[0017] like Figure 1-Figure 3 In this embodiment, a method for adaptively adjusting brightness of an LED backlight source may specifically include:

[0018] Step S101, using light sensors to obtain the intensity and spectral distribution information of downward natural light and indoor lighting sources in real time, superimpose the light sources, calculate the total light intensity and spectral distribution, and determine whether the adaptive adjustment conditions are met. If so, the LED backlight brightness adjustment process is triggered.

[0019] Collect indoor natural light intensity data and artificial light intensity data, calibrate and compensate the collected data according to optical measurement standards to obtain natural light parameter sets and artificial light parameter sets; perform discrete accumulation operation on spectral energy according to the natural light parameter set and the artificial light parameter set to obtain a total light intensity value, and use a spectral distribution curve superposition algorithm to obtain a total spectral distribution curve; compare the total light intensity value with a lighting threshold, and generate a backlight source adjustment trigger signal according to a matching result between the total spectral distribution curve and a spectral uniformity benchmark; control the backlight source lighting unit according to the backlight source adjustment trigger signal, collect total light parameters in a timed polling manner until the total light parameters meet the lighting threshold, and record the backlight source brightness parameters when the requirements are met and store them in a lighting parameter database.

[0020] Exemplarily, the optical sensor is used to collect the natural light intensity data and characteristic spectrum values ​​from the top of the room, and the artificial light source illumination intensity data and wavelength distribution values ​​are collected in the illumination area of ​​the indoor lamps. The collected data is calibrated and compensated according to the optical measurement standard to obtain the natural light parameter set and the artificial light parameter set. According to the natural light parameter set and the artificial light parameter set, the spectral energy in the wavelength range of 380-780 nanometers is discretely accumulated, and the total light intensity value is calculated using the linear superposition principle. The natural spectrum curve and the artificial spectrum curve are synthesized by the spectral distribution curve superposition method, and the total spectrum distribution model is established by the spectral energy density function, and the spectral energy distribution curve is output. The total light intensity value is compared with the standard lighting threshold, the spectral energy distribution curve is matched with the spectral uniformity benchmark, and the backlight adjustment trigger signal is generated. The backlight brightness is initialized and set by the light intensity mapping function, and the backlight brightness parameter is corrected by the luminous flux attenuation coefficient. The backlight lighting unit is controlled according to the corrected backlight brightness parameter, and the total light parameters are re-collected by a timed polling method until the light monitoring data meets the lighting threshold. The corresponding relationship between the backlight brightness parameters and the ambient light conditions when the lighting requirements are met is recorded and stored in the lighting parameter database. The optical sensor uses a high-precision photodiode array in the light data acquisition. The measurement wavelength range covers the visible spectrum range of 380-780 nanometers. The sampling interval is set to 1 nanometer. 25 measurement points are arranged in the sensor array, including 15 natural light monitoring points and 10 artificial light source monitoring points. The sampling frequency is 10 Hz. The output current signal of the photodiode is converted into a voltage signal through a transimpedance amplifier and digitized by a 16-bit analog-to-digital converter. The measured raw data needs to be dark current corrected and temperature compensated. In the spectral energy accumulation calculation, the spectral irradiance of each wavelength point is discretely summed. Assuming that the spectral irradiance at wavelength λ is E(λ), the total light intensity I is calculated as I=∑[E(λ)×Δλ], where Δλ is the wavelength interval. Taking a set of measured data as an example, the natural spectral irradiance measured at a wavelength of 500 nanometers is 0.85 watts / square meter / nanometer, and the spectral irradiance of the artificial light source is 0.65 watts / square meter / nanometer. After superposition calculation, the total spectral irradiance at this wavelength point is 1.50 watts / square meter / nanometer. The synthesis of the spectral distribution curve adopts the energy density weighted method, and the natural spectral curve and the artificial spectral curve are normalized and superimposed. Taking the measurement data at noon as an example, natural light is mainly distributed in the range of 450-650 nanometers, with a peak wavelength at 555 nanometers, and the spectral energy accounts for 60%; the artificial light source is mainly LED cold white light, with a bimodal distribution at 440 nanometers and 550 nanometers, and the spectral energy accounts for 40%.The backlight brightness control adopts a piecewise linear mapping method to divide the total light intensity into 5 gears: when the total light intensity is less than 100 lux, the backlight brightness is set to the maximum value; when the total light intensity is in the range of 100-300 lux, the backlight brightness is inversely proportional to the total light intensity; when the total light intensity is in the range of 300-500 lux, the backlight brightness is quadratically inversely proportional to the total light intensity; when the total light intensity is in the range of 500-800 lux, the backlight brightness decays logarithmically; when the total light intensity exceeds 800 lux, the backlight is turned off. The luminous flux attenuation coefficient takes into account the aging characteristics of LED devices. Through experimental measurements, it is found that the luminous flux of the LED decays to 95% of the initial value after 1000 hours of use, 92% after 2000 hours, and 90% after 3000 hours. Based on this, the corresponding relationship between LED life and luminous flux attenuation is established for real-time correction of the backlight brightness parameters. The corresponding relationship between the lighting parameters and backlight settings at different time periods and under different weather conditions is recorded in the lighting parameter database, so that the system can quickly respond to changes in ambient lighting.

[0021] Step S102, in the LED backlight brightness adjustment process, a spatial coupling model is established based on the indoor lighting structure, including the position and size of windows and skylights, as well as the layout position and angle of lamps, to analyze the factors affecting the diffuse distribution of natural light through the curtains, optimize the calculation method of the diffuse light intensity distribution, and obtain the ambient light intensity modulation coefficient of different areas.

[0022] The window position coordinate data collected by the ranging sensor and the curtain transmittance parameters measured by the photoelectric sensor are obtained, and a spatial coordinate set of indoor light source distribution is constructed according to the window position coordinate data and the curtain transmittance parameters; a spatial coupling relationship matrix of the lighting structure is established according to the curtain transmittance parameters in the spatial coordinate set, and the light transmission attenuation coefficient is calculated using the spatial coupling relationship matrix; the light transmission attenuation coefficient is calculated using a ray tracing algorithm, and a light intensity distribution diagram is generated through the spatial coupling relationship matrix; if the light intensity distribution diagram shows that the illumination of the key area is insufficient, the illumination compensation parameters of the key area are calculated, and a support vector regression algorithm is used to establish a mapping relationship between the compensation parameters and the ambient light, and a backlight area brightness modulation coefficient is output according to the mapping relationship.

[0023] Exemplarily, the indoor window position coordinate data and size parameters are obtained by scanning with a ranging sensor, the surface transmittance and diffuse reflection coefficient of the curtain are measured by a planar array photoelectric sensor, the angle of incident light from the skylight and the spatial position data of the lamp are collected, and the indoor light source distribution spatial coordinate set is constructed. According to the window transmittance data and the curtain diffuse reflection coefficient in the spatial coordinate set, the indoor lighting structure spatial coupling relationship matrix is ​​established, and the transmission attenuation coefficient of light between each lighting structure is calculated. The propagation path of natural light in the indoor space after being diffused by the curtain is calculated by a ray tracing algorithm, and the initial distribution map of light intensity is generated in combination with the spatial coupling relationship matrix. According to the initial distribution map of light intensity, the actual illumination value of each area is calculated by the regional integration method, and the spatial illumination distribution function is obtained by polynomial fitting. The illumination compensation parameters of the key indoor areas are calculated by the spatial illumination distribution function, and a corresponding relationship table between the compensation parameters and the backlight brightness is established. The compensation parameters are corrected by an iterative feedback method, and the mapping relationship between the compensation parameters and the ambient lighting conditions is established by a support vector regression algorithm. The backlight area brightness modulation coefficient is output according to the mapping relationship, and the corresponding relationship between the modulation coefficient and the ambient lighting parameter is recorded. The distance measuring sensor adopts the ultrasonic distance measuring principle, emits 40 kHz ultrasonic signals, calculates the propagation time by receiving the echo signal, and measures the position data of the window edge coordinate points. The window size is 2.4 meters wide, 1.8 meters high, and 1.2 meters from the ground. The array photoelectric sensor is composed of 16×16 silicon photodiodes. The measured curtain transmittance is 0.35, the diffuse reflection coefficient is 0.65, and the angle between the incident light of the skylight and the vertical direction is 30 degrees. The spatial coupling relationship of the indoor lighting structure is represented by a matrix. The matrix element aij represents the illumination contribution coefficient of the i-th lighting structure to the j-th area. Taking the 4×4 area division as an example, the contribution coefficient of the window to the window area is 0.8, to the second window area is 0.5, and to the far end area is 0.3; the contribution coefficient of the skylight to the area directly below is 0.9, and it decays outward to 0.7, 0.5, and 0.3 respectively. The ray tracing algorithm is based on the principle of geometric optics and decomposes the incident light into direct light and scattered light. Direct light follows the law of reflection, and scattered light conforms to Lambert's cosine law. When the light passes through the curtain, the incident light intensity is 100 lux. After passing through the curtain, the direct light intensity decays to 35 lux, and the scattered light is cosine distributed within the angle range of ±60 degrees. The spatial illumination distribution is fitted by a cubic polynomial, and the relationship between the illumination value E and the spatial position (x, y, z) is:

[0024] ,

[0025] E represents the illuminance value, x, y, z represent the spatial position coordinates, a, b, c, d, e, f, g, h, i, j are the unknown coefficients determined by the least squares method, and this formula describes the cubic polynomial relationship between the illuminance value and the spatial position. The measured data show that the illuminance in the window area is 850 lux, the illuminance in the middle of the room is 520 lux, and the illuminance in the far area is 320 lux. The illuminance compensation parameters are calculated based on the difference between the target illuminance and the actual illuminance. When the target illuminance is set to 500 lux, the compensation coefficient in the far area is 1.56, the compensation coefficient in the middle area is 0.96, and the compensation coefficient in the window area is 0.59. The support vector regression algorithm uses the radial basis kernel function, with kernel parameter γ=0.1 and relaxation factor C=10. The training samples contain illuminance data under different time periods and weather conditions. In the iterative feedback process, the actual illuminance value is remeasured after each adjustment of the backlight brightness, compared with the target illuminance, and the root mean square error is calculated. When the error is less than 30 lux, the iteration is stopped, and the brightness modulation coefficient of the backlight area is finally obtained. Experimental data shows that after 3-5 iterations, the illumination uniformity of each area in the room is improved to above 0.8, and the illumination standard deviation is reduced to below 50 lux.

[0026] Step S103, according to the spatial partition, the LED backlight source is divided into several independent units to be adjusted, the target brightness value of each unit to be adjusted is determined according to the ambient light intensity modulation coefficient, and the brightness of each unit to be adjusted is gradually adjusted using a smooth transition method to achieve the target brightness value.

[0027] The spatial distribution data is obtained by scanning the backlight area with an infrared sensor, and a quadtree partitioning algorithm is used to generate an adjustment unit number table for the backlight according to the spatial distribution data, and the adjustment unit number table records the spatial coordinate values ​​of the adjustment units; the initial brightness value of the overlapping part of the partition boundary is calculated by the adjacent area weighted average method according to the spatial coordinate values ​​of the adjustment units, and the target brightness sequence is obtained according to the ambient light intensity modulation coefficient; the brightness gradient curve of the adjustment unit is calculated for the target brightness sequence using a cubic spline interpolation algorithm, and a brightness gradient control sequence is generated according to the brightness gradient curve; the real-time brightness data of the adjustment unit is collected through a light intensity sensor array, and if there is a deviation value between the real-time brightness data and the target brightness sequence, a feedback neural network is used to construct a corresponding relationship between the brightness deviation and the compensation amount to obtain a brightness correction control signal.

[0028] Exemplarily, the backlight area is scanned by an infrared sensor to obtain spatial distribution data, the backlight is partitioned by a quadtree partitioning algorithm, a backlight adjustment unit number table is generated, and the spatial coordinate value and brightness detection reference value of each adjustment unit are recorded. The initial brightness value of the overlapping part of the partition boundary is calculated by the weighted average method of adjacent regions, and the target brightness sequence of each adjustment unit is generated according to the ambient light intensity modulation coefficient. The brightness gradient curve of each adjustment unit is calculated by the cubic spline interpolation algorithm, and the brightness gradient control sequence and timing control table are generated. Pulse width modulation is performed on each adjustment unit according to the timing control table, and the current brightness value of the adjustment unit is recorded by a linear buffer. The real-time brightness data of each adjustment unit is collected by an array of light intensity sensors, and the deviation value from the target brightness sequence is calculated. The corresponding relationship between the brightness deviation and the compensation amount is constructed by a feedback neural network, and a brightness correction control signal is generated. The pulse width modulation parameters are updated according to the brightness correction control signal, and the output drive voltage of each adjustment unit is dynamically adjusted. The backlight area is divided using a quadtree algorithm. The 16×12 cm backlight panel is recursively divided according to the brightness uniformity requirements. The first division forms 2×2 areas, the second division forms 4×4 areas, and finally forms 8×8 adjustment units. The area of ​​each adjustment unit is 2×1.5 cm. The infrared sensor scanning resolution is 0.1 mm. The center coordinates and boundary coordinates of each adjustment unit are recorded. The initial brightness detection reference value is set to 300 nits. In the weighted average processing of adjacent areas, the width of the overlapping area is set to 2 mm, and the distance inverse weight is used for calculation. If the brightness values ​​of two adjacent adjustment units are 280 nits and 320 nits respectively, the initial brightness value of the overlapping area is 280×0.6+320×0.4=296 nits. The ambient light intensity modulation coefficient ranges from 0.5 to 1.5. When the ambient light is dark, the modulation coefficient takes a larger value to enhance the brightness of the backlight. The cubic spline interpolation algorithm is used to generate a smooth brightness gradient curve. The interpolation interval is 100 milliseconds, and 20 interpolation points are calculated in each interval. Taking the brightness adjustment from 200 nits to 300 nits as an example, the cubic spline function is f(t)=at³+bt²+ct+d, where t is the normalized time variable, and the coefficients of a, b, c, and d are solved through boundary conditions, f(0)=200, f(1)=300, f'(0)=0, and f'(1)=0. The pulse width modulation adopts 12-bit resolution, the modulation frequency is 20 kHz, and the duty cycle step is 0.024%. The linear buffer capacity is 64 bytes, and the current brightness value is stored in the order of the adjustment unit number. The light intensity sensor array consists of 64 silicon photodiodes with a sensitivity of 0.5 volts / nit and a sampling frequency of 100 Hz. The feedback neural network adopts a three-layer structure. The input layer contains three neurons: the target brightness value, the current brightness value, and the ambient light intensity. The hidden layer uses 8 neurons, and the output layer is the brightness compensation amount.The training sample contains 1000 sets of brightness adjustment data. The network is trained using the Levenberg-Marquardt algorithm, and the root mean square error converges to within 5 nits. The brightness correction control adopts a proportional integral algorithm with a proportional coefficient Kp=0.8 and an integration time Ti=50 milliseconds. When the brightness deviation exceeds 10 nits, the correction control is triggered and the pulse width modulation parameters are updated. The driving voltage range is 0-5 volts, with a resolution of 10 millivolts, and is output through a DAC. Actual measured data shows that using this control method, the backlight brightness adjustment response time is less than 200 milliseconds, and the steady-state error is less than 3%.

[0029] Step S104, real-time detection of the actual brightness of each unit to be adjusted, comparison of the actual brightness with the optimized target brightness value, calculation of the brightness error, introduction of a machine learning algorithm to learn and predict the error data in real time, and dynamic adjustment of the PWM duty cycle of the unit to be adjusted.

[0030] The adjustment unit is sampled through a light intensity sensor array, and the sampled signal is converted into a digital signal and Gaussian filtered to obtain a light intensity data sequence; a linear mapping is performed based on the light intensity data sequence and a pre-established calibration curve to obtain a real-time brightness data sequence, and a difference operation is performed between the real-time brightness data sequence and the optimized target brightness sequence to obtain a brightness error sequence; the brightness error sequence is fitted using the recursive least squares method, and the fitted brightness error sequence is input into a convolutional neural network to generate a duty cycle prediction value based on the brightness error value and the ambient light conditions; the adjustment unit is driven by a pulse width modulator based on the duty cycle prediction value, the brightness response characteristic curve of the adjustment unit is collected, and the recursive least squares method parameters are updated based on the brightness response characteristic curve.

[0031] Exemplarily, each adjustment unit is sampled in real time through a light intensity sensor array, the collected light intensity data is converted into analog-to-digital data, and a Gaussian filter is used to eliminate sampling noise to generate a light intensity data sequence. The light intensity data sequence is linearly mapped according to the adjustment unit calibration curve to obtain a real-time brightness data sequence. The real-time brightness data sequence and the optimized target brightness sequence are differenced to generate a brightness error sequence. The brightness error sequence is fitted using the recursive least squares method to calculate the brightness error change trend. The brightness error sequence features are extracted using a convolutional neural network, the input layer includes the brightness error value and the ambient light conditions, the hidden layer extracts the timing features, and the output layer generates the duty cycle prediction value. Each adjustment unit is driven by a pulse width modulator, and the proportional integral differential operator dynamically adjusts the duty cycle. The adjusted brightness data is collected using a light intensity sensor array to generate a brightness response characteristic curve. The recursive least squares method parameters are updated according to the brightness response characteristic curve to optimize the duty cycle prediction function. The light intensity sensor array uses 16×16 silicon photodiodes with a sensitivity of 0.5 volts per nit and a sampling frequency of 100 Hz. The sampling signal is converted into a digital quantity by a 16-bit analog-to-digital converter, and the quantization accuracy reaches 0.01 nits. The Gaussian filter window size is set to 5, and the standard deviation is 1.2. The raw data is smoothed to effectively suppress high-frequency sampling noise. The calibration curve is obtained by measuring the standard light source. A calibration point is taken every 50 nits in the range of 0-500 nits. The mapping function is obtained by fitting a cubic polynomial, L=0.0023V³-0.0158V²+1.2456V+0.3251, where L is the brightness value and V is the sensor output voltage. In actual use, when the output voltage of a certain adjustment unit is 2.5 volts, the actual brightness value converted by the calibration curve is 285.6 nits. The recursive least squares method uses an exponential forgetting factor of 0.95 to fit the brightness error data of 10 consecutive sampling points. Taking a certain adjustment unit as an example, the brightness error values ​​of 5 consecutive sampling points were measured to be 8.5, 7.2, 6.1, 5.3, and 4.8 nits, respectively. The error change rate obtained by fitting is -0.92 nits / sampling cycle. The convolutional neural network structure contains 2 convolutional layers and 1 fully connected layer. The first convolutional layer uses 8 3×1 convolutional kernels to extract time series features, the second convolutional layer uses 16 3×1 convolutional kernels to extract high-order features, and the fully connected layer outputs the duty cycle prediction value. The network training uses 1000 sets of historical adjustment data, of which 70% is used for training and 30% is used for verification. After training, the root mean square error on the verification set is 0.8%. The pulse width modulator has an operating frequency of 20 kHz, a resolution of 12 bits, and a minimum adjustable duty cycle of 0.024%. In the proportional-integral-differential operation, the proportional coefficient Kp=0.8, the integral time Ti=50 milliseconds, and the differential time Td=10 milliseconds. When the brightness error is detected to be 10 nits, the duty cycle adjustment output is 2.5% after proportional-integral-differential calculation.The brightness response characteristic curve records the corresponding relationship between the duty cycle change and the brightness change, and is stored in a piecewise linear manner. Each adjustment unit stores 10 feature points. In actual applications, the current duty cycle of a certain adjustment unit is 40%, the brightness is 200 nits, and the target brightness is 250 nits. According to the response characteristic curve, the required duty cycle increment is 12%. The recursive least squares parameter update cycle is 1 second. The fitting parameters are dynamically adjusted according to the latest response characteristic curve to make the duty cycle prediction more accurate.

[0032] Step S105 , continuously obtaining user evaluation feedback data on display quality, performing cluster analysis on the evaluation feedback data through an optimized unsupervised learning algorithm, and generating a brightness adjustment curve adapted to different user preferences.

[0033] The real-time picture data of the display screen is obtained through an image collector, and the brightness distribution histogram and contrast value are extracted according to the real-time picture data to obtain a user brightness feedback data sequence; the principal component analysis method is used to extract brightness preference features according to the user brightness feedback data sequence, and the brightness preference features include average brightness value, brightness adjustment frequency, and brightness fluctuation range; the density peak clustering algorithm is used to perform unsupervised learning calculation on the brightness preference features to obtain feature space density distribution and brightness preference clustering center; a brightness preference feature and adjustment parameter mapping function is established for the brightness preference clustering center, and the cubic spline interpolation method is used to fit the brightness adjustment curve to obtain a piecewise linear brightness adjustment function.

[0034] Exemplarily, the real-time screen data of the display screen is obtained through an image collector, the screen brightness distribution histogram and contrast value are extracted, the user brightness adjustment operation record is recorded, and the user brightness feedback data sequence is generated. The user brightness feedback data sequence is preprocessed for data standardization to eliminate the dimension effect. The principal component analysis method is used to extract the brightness preference features of the standardized data, including three main feature dimensions: average brightness value, brightness adjustment frequency, and brightness fluctuation range. The density peak clustering algorithm is used to perform unsupervised learning on the brightness preference features and calculate the feature space density distribution. The brightness preference clustering center is determined according to the density distribution, and the user group division result is generated. A brightness preference feature and adjustment parameter mapping function is established for each group, and the brightness adjustment curve is fitted using the cubic spline interpolation method. A piecewise linear brightness adjustment function is generated according to the fitting curve, and the function parameters are recorded in the brightness adjustment database. The brightness adjustment function is tested using the cross-validation method to calculate the fitting error between the predicted brightness and the actual adjustment value. The image acquisition device is based on a CMOS image sensor with a resolution of 1920×1080 pixels and a sampling frequency of 60 Hz. The collected image data is statistically analyzed by a brightness histogram, and the brightness value is divided into 256 levels, and the number of pixels at each brightness level is recorded. The contrast value is obtained by calculating the ratio of the maximum brightness value to the minimum brightness value, and the typical value is between 300:1 and 1000:1. The user brightness adjustment record contains the adjustment timestamp, the brightness value before adjustment, and the brightness value after adjustment, and about 50-100 data are recorded every day. The Z-score method is used for data standardization, and the original data is returned to zero mean and normalized variance. Taking the brightness adjustment value as an example, the original data range is 0-255, the mean is 128, and the standard deviation is 42. The standardized data is distributed between -3 and 3. The three main feature dimensions extracted by principal component analysis explain 85%, 10%, and 3% of the data variance, respectively, and the cumulative explained variance reaches 98%. The density peak clustering algorithm calculates the local density ρi and the minimum distance δi of each data point, where ρi represents the number of points whose distance to data point i is less than the cutoff distance dc, and δi represents the distance to the nearest point with greater density. Points with larger ρi×δi values ​​are selected as cluster centers, and 3-5 user groups are usually formed in practical applications. The brightness preference characteristics of different user groups are significantly different. The first group prefers higher brightness, 200-250 nits, with low adjustment frequency, 1-2 times a day; the second group prefers medium brightness, 150-200 nits, with moderate adjustment frequency, 3-5 times a day; the third group prefers lower brightness, 100-150 nits, with high adjustment frequency, 6-8 times a day. Cubic spline interpolation uses a piecewise cubic polynomial function in each interval, f(x)=ax³+bx²+cx+d, and the coefficients are obtained by solving a system of linear equations.Taking the second user group as an example, the interpolation function coefficients in the 150-200 nits range are a=0.0012, b=-0.0458, c=1.2365, and d=85.6324. The piecewise linear brightness adjustment function discretizes the continuous curve into multiple linear intervals, and linear interpolation is used in each interval. For the second user group, the 150-200 nits range is divided into 5 segments, each with a length of 10 nits, and the starting brightness value and slope of each segment are stored. The cross-validation adopts a 5-fold method, randomly selecting 80% of the data for training and 20% of the data for testing, with an average prediction error of less than 5 nits. In the application scenario, when a new user initially uses it, features are extracted based on their first 10 brightness adjustment behaviors, and similarity is calculated with the existing group features to match the closest user group and use the corresponding brightness adjustment function. As the user's usage time increases, their brightness preference features are continuously updated, and group affiliation and adjustment function parameters are dynamically adjusted.

[0035] Step S106, based on the dynamic change characteristics of the composite light field, a prediction model of total light intensity, spectral distribution, and spatial distribution is established through a machine learning algorithm, and the light field change trend within a preset time period is estimated in advance to obtain a composite light field prediction result, and the prediction result is used to adaptively optimize the adjustment strategy of the backlight source.

[0036] The raw data sequence of indoor light field is obtained through a photoelectric sensor array, and the raw data sequence of the light field includes light intensity value, wavelength distribution parameter and spatial coordinate information; a light field time series feature matrix is ​​generated by a sliding window method according to the raw data sequence of the light field, and the sliding window covers multiple sampling period data; a long short-term memory neural network model is established for the light field time series feature matrix, and the input of the neural network model includes historical light intensity value and environmental parameters; a Kalman filter is constructed according to the output result of the neural network model, and the state vector of the Kalman filter includes light intensity prediction value, main wavelength component and spatial feature data, and a variational autoencoder is used to perform online optimization on the output of the Kalman filter.

[0037] Exemplarily, indoor composite light field data is obtained through a photoelectric sensor array, light field parameters including light intensity value, wavelength distribution, and spatial coordinates are collected, and an original light field data sequence is generated. The original light field data sequence is segmented by a sliding window, and the window length covers multiple sampling periods to generate a light field time series feature matrix. A light intensity prediction sub-model is established using a long short-term memory neural network, and the input features include historical light intensity values ​​and environmental parameters, and the hidden layer extracts time series features. The principal component analysis method is used to reduce the dimension of the spectral distribution data and extract the main wavelength feature vector. The spatial distribution data is normalized to the maximum and minimum values ​​to generate a standardized spatial feature matrix. A Kalman filter is constructed based on the light intensity prediction value, wavelength feature vector, and spatial feature matrix, and the state vector includes light intensity, main wavelength component, and spatial coordinates. The Kalman filter is used to recursively predict the light field state within a preset time period to generate a light field change trend curve. The prediction results are optimized online through a variational autoencoder, and the backlight brightness adjustment parameters are output, and the prediction error is recorded for model updating. The photoelectric sensor array consists of 8×8 silicon photodiodes, each with an effective collection area of ​​1 square millimeter, a sensitivity of 0.5 volts / nit, and a sampling frequency of 100 Hz. The light field parameters include light intensity values ​​0-1000 nits, wavelength distribution 380-780 nanometers, and spatial coordinates (a three-dimensional coordinate system with the center of the sensor array as the origin). The sliding window length is set to 100 sampling points, and the sliding step is 10 sampling points to segment the continuous sampling data. Taking the light intensity data as an example, the data in the window forms a 100×64 time series feature matrix, each row corresponds to a time point, and each column corresponds to a sensor. Environmental parameters include temperature 15-35 degrees Celsius and humidity 30%-70%. The long short-term memory neural network adopts a three-layer structure, with 66 neurons in the input layer, 64 light intensity values ​​plus 2 environmental parameters, 128 memory units in the hidden layer, and the output layer predicts the light intensity values ​​at the next 10 time points. The Adam optimizer was used for network training, with a learning rate of 0.001 and a batch size of 32. Principal component analysis was used to reduce the dimension of the wavelength distribution data. The original data contained 401 wavelength points, 380-780 nanometers, with an interval of 1 nanometer. The first five principal components were extracted, and the cumulative explained variance reached 95%. The first principal component reflects the overall light intensity change, the second and third principal components characterize the spectral peak position, and the fourth and fifth principal components describe the spectral shape. The spatial feature normalization uses the maximum and minimum method. The three coordinates x, y, and z are divided by their respective coordinate ranges and mapped to the 0-1 interval. The dimension of the normalized spatial feature matrix is ​​64×3, and each row represents the standardized spatial position of a sensor. The state vector of the Kalman filter contains 8 components: light intensity value, light intensity change rate, 3 main wavelength components, and 3 spatial coordinates. The state transfer matrix uses a uniform motion model, and the observation matrix maps the state vector to the actual measurement value.The diagonal elements of the process noise covariance matrix are set to 0.1, and the diagonal elements of the measurement noise covariance matrix are set to 0.05. The encoder and decoder of the variational autoencoder both use a three-layer fully connected network with a hidden layer dimension of 16 and a latent space dimension of 8. The input feature is the predicted light field state vector, and the output is the optimized backlight control parameter. The training uses the weighted sum of the reconstruction error and KL divergence as the loss function, with weight coefficients of 0.8 and 0.2 respectively. The root mean square error of the prediction result is controlled within 5%, and the model updates the parameters every 100 sampling points.

[0038] Step S107, based on the multi-light sensor data collected in real time, composite light field prediction is performed, and the spatial brightness distribution of the current display screen is used as the optimization objective function. The PWM duty cycle of each unit to be adjusted of the backlight source is solved in real time through the reinforcement learning algorithm to obtain the dynamically optimized backlight brightness adjustment strategy, which is transmitted to the LED backlight drive circuit to execute the corresponding brightness adjustment action.

[0039] A real-time light intensity data sequence of a displayed image captured by a multi-light sensor array is obtained, and a pixel brightness distribution matrix is ​​generated according to the light intensity data sequence; a mean square error optimization objective function is constructed according to the pixel brightness distribution matrix, and a gradient value of the objective function is obtained by using a gradient descent method; a state vector of an adjustment unit is constructed according to the gradient value, wherein the state vector includes a light intensity value, a position coordinate, and a duty cycle parameter, and a reward value is calculated according to the state vector and the objective function; a gradient strategy method is used to iteratively calculate the reward value to obtain a duty cycle adjustment amount, and a backlight control sequence is generated according to the duty cycle adjustment amount, and after acquiring an actual brightness response curve, the duty cycle adjustment amount is corrected according to a response deviation value.

[0040] Exemplarily, a real-time light intensity data sequence is obtained through a multi-light sensor array, and a pixel brightness distribution matrix of the current display screen is collected to generate a light field measured data set. The prediction accuracy is calculated based on the mean square error between the light field measured data set and the predicted value, and the parameters of the composite light field prediction function are corrected. The pixel brightness distribution matrix is ​​used to construct a mean square error optimization objective function, and the objective function gradient is calculated. The state vector of the adjustment unit is constructed using deep reinforcement learning, which includes light intensity value, position coordinates and duty cycle parameters. The reward value is calculated based on the state vector and the optimization objective function, and a corresponding relationship between the reward value and the duty cycle adjustment amount is established. The duty cycle adjustment amount is iteratively optimized by the gradient strategy method to generate a backlight control sequence. A pulse modulation signal is applied to the backlight adjustment unit to collect the actual brightness response curve. The duty cycle adjustment amount is corrected online according to the response curve deviation value, and the reinforcement learning parameters are updated. The multi-light sensor array is composed of 8×8 photodiodes, with a sampling frequency of 100 Hz, a sensitivity of 0.5 V / nit, and a quantization accuracy of 12 bits. The pixel brightness distribution matrix uses a resolution of 1920×1080, and the brightness value of each pixel ranges from 0 to 255. In actual tests, the measured data of the light field contains the light intensity values ​​of 64 sampling points, which are distributed in a 16×9 display area, and the distance between adjacent sampling points is 10 cm. The prediction accuracy is calculated using the root mean square error indicator. Taking a certain set of data as an example, the predicted light intensity value sequence is [450,460,455,465] nits, and the measured value is [460,455,465,470] nits. The calculated root mean square error is 7.07 nits, and the prediction accuracy is 98.5%. The optimization objective function uses the pixel brightness mean square error form, f(x)=Σ(xi-μ)² / n, where xi is the brightness value of the i-th pixel, μ is the average brightness value, and n is the total number of pixels. The state vector dimension of deep reinforcement learning is 66, including 64 light intensity measurement values ​​and 2 position coordinate values. The action space of each adjustment unit is the duty cycle increment value, ranging from [-5%, +5%], discretized into 21 levels. The reward function is designed as R=-w1×MSE-w2×VAR, where MSE is the mean square error of brightness, VAR is the brightness variance, and the weight coefficients w1=0.7 and w2=0.3. The gradient strategy method uses the stochastic gradient ascent algorithm, and the initial value of the learning rate is set to 0.01, which gradually decays with the training rounds. Taking a certain adjustment unit as an example, the light intensity value in the current state is 400 nits, and the target light intensity value is 450 nits. After 5 rounds of iterative calculation, the optimal duty cycle increment is +3.2%. The duty cycle modulation of the backlight adjustment unit adopts a 20 kHz carrier frequency, a resolution of 12 bits, and a minimum adjustment step of 0.024%. The brightness response curve records the corresponding relationship between the duty cycle change and the actual brightness change, and is stored in a piecewise linear manner. In actual applications, when the duty cycle of a certain adjustment unit increases from 40% to 43.2%, the brightness increases from 400 nits to 442 nits, which is lower than the expected 450 nits.According to the response curve deviation, the compensation coefficient of 1.12 is calculated, and the corrected duty cycle increment is updated to 3.6%. The reinforcement learning parameter update adopts the experience replay mechanism, the experience pool capacity is 1000 groups, and 32 groups of data are randomly sampled for training each time. The value network adopts a three-layer fully connected structure, the hidden layer dimension is 128, the activation function is ReLU, and the training adopts the Adam optimizer. Through online learning, the policy network gradually adapts to the nonlinear characteristics of the backlight source, and the duty cycle prediction accuracy is increased from the initial 85% to more than 95%.

[0041] It should be noted that the above examples are only some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.

Claims

1. A method for adaptively adjusting brightness of an LED backlight source, characterized in that: The method comprises: obtaining the intensity and spectral distribution information of downward natural light and indoor lamp light sources in real time through a light sensor, superimposing the light sources, calculating the total light intensity and spectral distribution, judging whether the adaptive adjustment conditions are met, and triggering the LED backlight source brightness adjustment process if the conditions are met; in the LED backlight source brightness adjustment process, establishing a spatial coupling model according to the indoor lighting structure, including the position and size of windows and skylights, and the arrangement position and angle of lamps, analyzing the influencing factors on the diffuse distribution of natural light after passing through the curtains, optimizing the calculation method of the diffuse light intensity distribution, and obtaining the ambient light intensity modulation coefficient of different areas; dividing the LED backlight source into a plurality of independent units to be adjusted according to the spatial partition, determining the target brightness value of each unit to be adjusted according to the ambient light intensity modulation coefficient, and gradually adjusting the brightness of each unit to be adjusted by a smooth transition method to achieve the target brightness value; detecting the actual brightness of each unit to be adjusted in real time, and comparing the actual brightness with the target brightness value. Yes, calculate the brightness error, introduce a machine learning algorithm to learn and predict the error data in real time, and dynamically adjust the PWM duty cycle of the unit to be adjusted; continuously obtain user evaluation feedback data on display quality, perform cluster analysis on the evaluation feedback data through an optimized unsupervised learning algorithm, and generate a brightness adjustment curve that adapts to the preferences of different users; based on the dynamic change characteristics of the composite light field, establish a prediction model of total light intensity, spectral distribution, and spatial distribution through a machine learning algorithm, and predict the light field change trend within a preset time period in advance to obtain the composite light field prediction result, and use the prediction result to adaptively optimize the backlight adjustment strategy; perform composite light field prediction based on real-time collected multi-light sensor data, and use the spatial brightness distribution of the current display screen as the optimization objective function, and use the reinforcement learning algorithm to solve the PWM duty cycle of each unit to be adjusted in the backlight source in real time to obtain a dynamically optimized backlight brightness adjustment strategy, which is transmitted to the LED backlight drive circuit to perform the corresponding brightness adjustment action.

2. The method according to claim 1, characterized in that The light sensor is used to respectively obtain the intensity and spectral distribution information of downward natural light and indoor lamp light sources in real time, perform light source superposition, calculate the total light intensity and spectral distribution, determine whether the adaptive adjustment conditions are met, and trigger the LED backlight brightness adjustment process if the conditions are met, including: collecting indoor natural light intensity data and artificial light source illumination intensity data, calibrating and compensating the collected data according to optical measurement standards to obtain a natural light parameter set and an artificial light parameter set; according to the natural light parameter set and the artificial light parameter set, performing discrete accumulation operation on spectral energy to obtain a total light intensity, and using a spectral distribution curve superposition algorithm to obtain a total spectral distribution curve; comparing the total light intensity with an illumination threshold, and generating a backlight adjustment trigger signal according to a matching result between the total spectral distribution curve and a spectral uniformity benchmark; controlling the backlight illumination unit according to the backlight adjustment trigger signal, collecting the total light intensity in a timed polling manner until the total light intensity meets the illumination threshold, and recording the backlight brightness parameters when the requirements are met and storing them in an illumination parameter database.

3. The method according to claim 1, characterized in that In the LED backlight brightness adjustment process, according to the indoor lighting structure, including the position and size of windows and skylights, and the arrangement position and angle of lamps, a spatial coupling model is established, the influencing factors on the diffuse distribution of natural light after passing through the curtains are analyzed, the calculation method of the diffuse light intensity distribution is optimized, and the ambient light intensity modulation coefficient of different areas is obtained, including: obtaining the window position coordinate data collected by the ranging sensor and the curtain transmittance parameters measured by the photoelectric sensor, and constructing an indoor light source distribution spatial coordinate set according to the window position coordinate data and the curtain transmittance parameters; establishing a lighting structure spatial coupling relationship matrix according to the curtain transmittance parameters in the spatial coordinate set, and calculating the light transmission attenuation coefficient by using the spatial coupling relationship matrix; using a ray tracing algorithm to calculate the light transmission attenuation coefficient, and generating a light intensity distribution map through the spatial coupling relationship matrix; if the light intensity distribution map shows that the illumination of the key area is insufficient, then calculating the illumination compensation parameters of the key area, using a support vector regression algorithm to establish a mapping relationship between the compensation parameters and the ambient light, and outputting the ambient light intensity modulation coefficients of different areas according to the mapping relationship.

4. The method according to claim 1, characterized in that: According to the spatial partition, the LED backlight source is divided into a number of independent units to be adjusted, the target brightness value of each unit to be adjusted is determined according to the ambient light intensity modulation coefficient, and the brightness of each unit to be adjusted is gradually adjusted by a smooth transition method to achieve the target brightness value, including: scanning the backlight source area by an infrared sensor to obtain spatial distribution data, generating an adjustment unit number table for the backlight source by a quadtree partitioning algorithm according to the spatial distribution data, and the adjustment unit number table records the spatial coordinate values ​​of the adjustment units; calculating the initial brightness value of the overlapping part of the partition boundary by a weighted average method of adjacent areas according to the spatial coordinate values ​​of the adjustment units, and obtaining the target brightness sequence according to the ambient light intensity modulation coefficient; calculating the brightness gradient curve of the adjustment unit for the target brightness sequence by a cubic spline interpolation algorithm, and generating a brightness gradient control sequence according to the brightness gradient curve; collecting real-time brightness data of the adjustment unit by a light intensity sensor array, and if there is a deviation value between the real-time brightness data and the target brightness sequence, using a feedback neural network to construct a corresponding relationship between the brightness deviation and the compensation amount to obtain a brightness correction control signal.

5. The method according to claim 1, characterized in that The method of detecting the actual brightness of each unit to be adjusted in real time, comparing the actual brightness with the optimized target brightness value, calculating the brightness error, introducing a machine learning algorithm to learn and predict the error data in real time, and dynamically adjusting the PWM duty cycle of the unit to be adjusted includes: sampling the adjustment unit through a light intensity sensor array, obtaining a light intensity data sequence after analog-to-digital conversion and Gaussian filtering of the sampled signal; performing linear mapping according to the light intensity data sequence and a pre-established calibration curve to obtain a real-time brightness data sequence, performing a difference operation between the real-time brightness data sequence and the optimized target brightness sequence to obtain a brightness error sequence; The brightness error sequence is fitted by using a recursive least square method, the brightness error sequence after fitting is input into a convolutional neural network, and a duty cycle prediction value is generated according to the brightness error value and the ambient light condition; The regulating unit is driven according to the duty cycle prediction value by a pulse width modulator, a brightness response characteristic curve of the regulating unit is collected, and the recursive least square method parameters are updated according to the brightness response characteristic curve.

6. The method according to claim 1, characterized in that The method comprises: continuously acquiring user evaluation feedback data on display quality, performing cluster analysis on the evaluation feedback data through an optimized unsupervised learning algorithm, and generating a brightness adjustment curve adapted to different user preferences, including: acquiring real-time screen data of the display screen through an image collector, extracting a brightness distribution histogram and a contrast value according to the real-time screen data, and obtaining a user brightness feedback data sequence; extracting brightness preference features according to the user brightness feedback data sequence by using a principal component analysis method, wherein the brightness preference features include an average brightness value, a brightness adjustment frequency, and a brightness fluctuation range; and performing unsupervised learning calculation on the brightness preference features by using a density peak clustering algorithm to obtain a feature space density distribution and a brightness preference clustering center; A brightness preference feature and adjustment parameter mapping function is established for the brightness preference cluster center, and a cubic spline interpolation method is used to fit the brightness adjustment curve to obtain a piecewise linear brightness adjustment function.

7. The method according to claim 1, characterized in that The method aims at the dynamic change characteristics of the composite light field, establishes a prediction model of total light intensity, spectral distribution, and spatial distribution through a machine learning algorithm, predicts the change trend of the light field in a preset time period in advance, so as to obtain a composite light field prediction result, and uses the prediction result for adaptively optimizing the adjustment strategy of the backlight source, including: obtaining an indoor light field raw data sequence through a photoelectric sensor array, wherein the light field raw data sequence includes light intensity values, wavelength distribution parameters, and spatial coordinate information; generating a light field time series feature matrix according to the light field raw data sequence using a sliding window method, wherein the sliding window covers multiple sampling period data; establishing a long short-term memory neural network model for the light field time series feature matrix, wherein the neural network model input includes historical light intensity values ​​and environmental parameters; constructing a Kalman filter according to the output result of the neural network model, wherein the Kalman filter state vector includes light intensity prediction values, main wavelength components, and spatial feature data, and online optimizing the Kalman filter output using a variational autoencoder.

8. The method according to claim 1, characterized in that The composite light field prediction is performed based on the multi-light sensor data collected in real time, and the spatial brightness distribution of the current display screen is used as the optimization objective function. The PWM duty cycle of each unit to be adjusted of the backlight source is solved in real time through the reinforcement learning algorithm to obtain the dynamically optimized backlight brightness adjustment strategy, which is transmitted to the LED backlight driving circuit to perform the corresponding brightness adjustment action, including: obtaining the real-time light intensity data sequence of the displayed screen collected by the multi-light sensor array, and generating a pixel brightness distribution matrix according to the light intensity data sequence; constructing a mean square error objective function according to the pixel brightness distribution matrix, and using the gradient descent method to obtain the gradient value of the objective function; constructing an adjustment unit state vector according to the gradient value, the state vector including the light intensity value, the position coordinates and the duty cycle parameter, and calculating the reward value according to the state vector and the objective function; The reward value is iteratively calculated using a gradient strategy method to obtain a duty cycle adjustment amount, a backlight source control sequence is generated according to the duty cycle adjustment amount, and the duty cycle adjustment amount is corrected according to a response deviation value after an actual brightness response curve is collected.

Images