AI-based LED lamp bead brightness self-adaptive adjustment method and system

Abstract

This invention relates to the field of LED lighting technology, and more particularly to an AI-based method and system for adaptive brightness adjustment of LED chips. By performing spatiotemporal decoupling and colorimetric analysis on multidimensional spectral data, the method extracts illumination fluctuation characteristics and color fidelity characteristic sequences. These sequences are converted into pulse-coded forms and coupled temporally via synaptic delay chains and lateral suppression mechanisms to obtain an illumination fit evaluation tensor. Based on the three-dimensional geometric structure, the method calculates the illumination contribution path and attenuation coefficient, combining the evaluation tensor to generate a spatial illuminance demand distribution matrix. Theoretical contribution matrices are calculated based on the parameters of each LED chip, and a performance attenuation compensation mechanism is introduced. Iterative solutions are used to approximate the demand distribution matrix with the product, and the theoretical contribution matrix is ​​dynamically adjusted. After convergence, the final driving intensity vector is obtained and converted into an adjustment command. This invention achieves precise adaptive brightness adjustment, improving color reproduction quality and illumination uniformity.

Description

Technical Field

[0001] This invention relates to the field of LED lighting technology, and in particular to an AI-based method and system for adaptive brightness adjustment of LED beads. Background Technology

[0002] In current LED lighting control technology, adaptive brightness adjustment methods generally rely on sensor feedback or preset scene modes. They collect basic parameters such as illuminance and color temperature within the space and combine this with PWM duty cycle adjustment or constant current drive to perform closed-loop control of the LEDs. Conventional approaches typically focus on maintaining the target illuminance level, reducing energy consumption, or achieving gradual color temperature changes as core optimization goals. For example, they adjust output brightness in real-time based on ambient light sensors or use infrared sensors to detect the location of people and then perform zoned dimming. These methods can meet basic lighting needs in static or slowly changing scenarios, but their adjustment performance is significantly limited in dynamic and complex lighting environments, such as spaces with moving objects, frequent changes in natural light, or diverse surface reflectivity.

[0003] Traditional methods focus only on the physical intensity or color temperature of light, lacking the ability to analyze the dynamic coupling relationship between light fluctuations and the color response of object surfaces. Specifically, when lighting conditions change rapidly (such as a sudden change in brightness in a window area due to cloud cover), the chromaticity information reflected from the object surface undergoes nonlinear distortion. Conventional control algorithms cannot assess the impact of this distortion on color fidelity in real time, leading to a decrease in the color reproduction quality of the illuminated object, a problem that is particularly prominent in scenarios where extremely high color accuracy is required.

[0004] Conventional methods neglect the performance degradation differences between individual LED chips. Over time, the luminous flux of an LED chip gradually decreases due to accumulated junction temperature and phosphor aging. Different chips exhibit inconsistent degradation rates, causing the actual illuminance distribution in the space to deviate from the initial design, even when using the same drive commands. Existing linear compensation models typically perform coarse calibration based on average lifetime curves, failing to dynamically correct for the historical operating state of each chip. This leads to a gradual deterioration of the system's adjustment accuracy over long-term use, requiring frequent manual calibration or chip replacement, increasing maintenance costs. Therefore, there is an urgent need for an intelligent adjustment scheme that can balance color reproduction quality in dynamic lighting environments with adapting to the individual chip degradation. Summary of the Invention

[0005] This invention provides an AI-based method and system for adaptive brightness adjustment of LED beads, which can solve the problems in the prior art.

[0006] A first aspect of this invention provides an AI-based method for adaptive brightness adjustment of LED beads, comprising: Spatiotemporal decoupling analysis was performed on the multidimensional spectral distribution data within the lighting space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis was conducted on the reflectance spectral characteristic data to obtain a color fidelity feature sequence. The illumination fluctuation characteristics and the color fidelity characteristic sequences are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms to obtain an illumination fit evaluation tensor that characterizes the object's color reproduction quality under the current illumination environment. Based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial position index in the lighting adaptability evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated. The theoretical contribution matrix of each LED to each point in space is calculated based on the spatial position and beam angle parameters of each LED. The initial normalized driving intensity vector of each LED is set, and a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED is introduced. The product of the theoretical contribution matrix and the initial normalized driving intensity vector is iteratively solved so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

[0007] Spatiotemporal decoupling analysis was performed on the multidimensional spectral distribution data within the lighting space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis was conducted on the reflectance spectral characteristic data to obtain the color fidelity feature sequence, including: The multidimensional spectral distribution data, which includes spatial distribution information and temporal evolution information of different wavelength components within the visible light band, is collected within the illumination space using an array of photoelectric sensors. The time evolution information is subjected to multi-level wavelet decomposition. During the decomposition process, wavelet coefficients at each scale are extracted and cross-correlation analysis is performed to identify cross-scale coupled illumination fluctuation patterns. Spatial coherence analysis is performed on the spatial distribution information to obtain the synchronization and phase difference of illumination fluctuations between different sampling locations as spatial coupling features. The spatial coupling features are then fused with the cross-scale coupled illumination fluctuation patterns to obtain illumination fluctuation features at different scales that include spatiotemporal coupling relationship identifiers. The reflectance spectral characteristic data of the object surface in the illumination space under the current illumination conditions are collected by a distributed spectrometer. The reflectance spectral characteristic data are converted into the chromaticity space to obtain a measured chromaticity coordinate sequence. The color difference value sequence between the measured chromaticity coordinate sequence and the reference chromaticity coordinate sequence of the object surface under the standard illuminator is calculated. Each color difference value in the color difference value sequence is normalized and compared with the human eye color perception threshold to obtain the color fidelity characteristic sequence that characterizes the accuracy of color reproduction of the object surface.

[0008] The illumination fluctuation characteristics and the color fidelity characteristic sequences are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms. This yields an illumination fit evaluation tensor characterizing the object's color reproduction quality under the current illumination environment, including: The spatiotemporal coupling relationship identifier in the illumination fluctuation feature is mapped to the pulse emission spatiotemporal code, and the color fidelity feature sequence is mapped to the pulse emission frequency sequence; A synaptic delay chain containing multiple parallel delay paths is constructed. Each delay path corresponds to a time scale in the illumination fluctuation characteristics at different scales. The delay time parameters of each delay path are set according to the pulse emission spatiotemporal coding, so that the pulse signal in the pulse emission spatiotemporal coding generates a time delay effect corresponding to the illumination change rate when it propagates in the synaptic delay chain. A lateral inhibition mechanism is constructed, in which the inhibitory neurons dynamically adjust the lateral inhibition intensity according to the pulse firing frequency sequence. The lateral inhibition intensity is inversely proportional to the pulse frequency in the pulse firing frequency sequence, thereby realizing the encoding expression of the color response rate of the object surface. The output pulse sequence of the synaptic delay chain is time-aligned with the output pulse sequence of the lateral inhibition mechanism and the temporal correlation is calculated. The temporal correlation characterizes the temporal coupling relationship between the rate of change of illumination and the color response rate of the object surface. The illumination fit evaluation tensor is generated based on the temporal coupling relationship, and the illumination fit evaluation tensor includes the spatial location index in the pulse firing spatiotemporal coding as a spatial location index.

[0009] Based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial location index in the lighting fit evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated, including: Based on the spatial coordinate information and surface normal vector information in the three-dimensional geometric structure data, the reflectivity parameters and diffuse reflection direction distribution parameters of each surface are determined. Starting from the installation position of the LED beads, a tracking light is emitted to each surface. At each intersection point, a reflected light is generated according to the reflectivity parameter and the diffuse reflection direction distribution parameter. The propagation path of the light is iteratively tracked multiple times. The illumination contribution corresponding to each light propagation path is calculated and accumulated to obtain the total illumination contribution path of each surface. The multiple reflection attenuation coefficient is calculated based on the number of reflections in the light propagation path. The multiple reflection attenuation coefficient is equal to the product of the reflectivity parameters of each reflection in the light propagation path. Spatial matching is performed between the spatial location index in the illumination adaptability evaluation tensor and the spatial coordinate information to determine the surface position corresponding to each spatial location index. The illuminance requirement is calculated based on the total light contribution path at the surface location and the multiple reflection attenuation coefficient. The illuminance requirement is equal to the product of the total light contribution path and the reciprocal of the multiple reflection attenuation coefficient. The position priority weight is calculated based on the fitness value corresponding to each spatial position index in the illumination fitness evaluation tensor. The position priority weight is proportional to the fitness value. The illumination demand value and the position priority weight are combined to generate the spatial illumination demand distribution matrix.

[0010] Based on the spatial position and beam angle parameters of each LED, the theoretical contribution matrix of each LED to each point in space is calculated. An initial normalized driving intensity vector for each LED is set. A performance degradation compensation mechanism based on the historical operating time and junction temperature accumulation of the LEDs is introduced, including: The beam angle parameters of each LED bead include the beam divergence angle and the spatial distribution characteristics of light intensity; Calculate the distance vector from each LED to each spatial point in the spatial illuminance demand distribution matrix based on the spatial position coordinates of each LED. Determine whether each spatial point is within the effective irradiation range of each LED based on the distance vector and the beam divergence angle. For spatial points within the effective irradiation range, calculate the theoretical illuminance contribution value of each LED to each spatial point based on the length of the distance vector and the light intensity spatial distribution characteristics, and organize it into a theoretical contribution matrix. Set the initial normalized driving intensity vector for each LED; The historical operating time data and junction temperature accumulation data of each LED bead are obtained. The operating time attenuation factor is calculated based on the historical operating time data. The operating time attenuation factor is inversely proportional to the historical operating time data. The temperature attenuation factor is calculated based on the junction temperature accumulation data. The temperature attenuation factor is inversely proportional to the junction temperature accumulation data. The comprehensive performance attenuation coefficient is obtained by multiplying the operating time attenuation factor and the temperature attenuation factor.

[0011] Iteratively solving for the product of the theoretical contribution matrix and the initial normalized driving intensity vector, so that the product approximates the spatial illuminance demand distribution matrix, includes: The initial illuminance distribution prediction value is obtained by performing matrix multiplication between the theoretical contribution matrix and the initial normalized driving intensity vector, and the illuminance difference matrix between the initial illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​calculated. The gradient direction is calculated on the initial normalized driving intensity vector based on the illuminance difference matrix. The gradient direction points in the direction that reduces the total error of the illuminance difference matrix. The initial normalized driving intensity vector is updated along the gradient direction to obtain an intermediate normalized driving intensity vector. Elements in the intermediate normalized driving intensity vector that are outside the normalized value range are subjected to projection constraints so that all elements remain within the normalized value range to obtain the constrained normalized driving intensity vector.

[0012] During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads, including: During the iteration process, the row vectors corresponding to each LED bead in the theoretical contribution matrix are scaled and adjusted according to the comprehensive performance attenuation coefficient. The scaling ratio of the scaling adjustment is equal to the comprehensive performance attenuation coefficient, thus obtaining the dynamically adjusted theoretical contribution matrix. The updated illuminance distribution prediction value is obtained by performing matrix multiplication between the dynamically adjusted theoretical contribution matrix and the constrained normalized driving intensity vector. The updated illuminance difference matrix between the updated illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​then calculated. Determine whether the total error of the updated illuminance difference matrix meets the convergence condition. The convergence condition is that the total error of the updated illuminance difference matrix is ​​less than a preset error threshold or the number of iterations reaches a preset maximum number of iterations. If the convergence condition is met, the constrained normalized driving intensity vector is determined as the final normalized driving intensity vector. If the convergence condition is not met, the constrained normalized driving intensity vector is used as the new initial normalized driving intensity vector and the gradient direction is recalculated. Based on the normalized drive intensity value of each LED in the final normalized drive intensity vector, an adjustment command is generated for the corresponding LED, and the adjustment command is applied to the drive circuit of the corresponding LED.

[0013] A second aspect of this invention provides an AI-based adaptive brightness adjustment system for LED beads, comprising: The spectral analysis unit is used to perform spatiotemporal decoupling analysis on multidimensional spectral distribution data in the lighting space, extract illumination fluctuation characteristics at different scales, perform colorimetric analysis on reflectance spectral characteristic data, and obtain color fidelity characteristic sequences. The temporal coupling unit is used to convert the light fluctuation characteristics and the color fidelity characteristic sequence into pulse code form, and establish the temporal coupling relationship between the light change rate and the object surface color response rate through synaptic delay chain and lateral inhibition mechanism to obtain the light fit evaluation tensor characterizing the color reproduction quality of the object under the current lighting environment. The path requirement unit is used to calculate the light contribution path and multiple reflection attenuation coefficient of each surface in the space based on the three-dimensional geometric structure data of the lighting space, and generate a spatial illuminance requirement distribution matrix by combining the spatial position index in the light fit evaluation tensor. The drive compensation unit is used to calculate the theoretical contribution matrix of each LED to each point in space based on the spatial position and beam angle parameters of each LED, set the initial normalized drive intensity vector of each LED, and introduce a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED. The iterative driving unit is used to iteratively solve the product of the theoretical contribution matrix and the initial normalized driving intensity vector, so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance decay compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

[0014] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] The beneficial effects of this application are as follows: The light fluctuation features and color fidelity feature sequences extracted by spatiotemporal decoupling analysis provide a multi-scale data foundation for accurately assessing the impact of light on color reproduction. It can effectively eliminate color distortion caused by dynamic changes in ambient light, significantly improve the authenticity and stability of color reproduction on object surfaces in complex lighting scenarios, and enable the lighting system to perceive subtle changes in spectral distribution, thereby actively maintaining color balance during dimming.

[0017] By combining pulse coding with synaptic delay chain and lateral inhibition mechanism to establish the temporal coupling relationship between illumination change rate and color response rate, an illumination adaptability evaluation tensor characterizing the spatiotemporal distribution of color reproduction quality is generated. This tensor can reflect the differentiated needs of illumination adaptability at various locations in the lighting space in real time, providing a directional optimization basis for subsequent LED bead brightness adjustment and reducing visual discomfort and color deviation caused by response lag.

[0018] The spatial illuminance demand distribution matrix is ​​combined with the theoretical contribution matrix of LED chips and the performance degradation compensation mechanism in an iterative solution process. In the successive approximation, the output deviation caused by chip aging and junction temperature accumulation is dynamically corrected. Finally, the normalized drive intensity vector ensures that the lighting system achieves uniform and high color fidelity illuminance distribution with the lowest energy consumption, while extending the working life of LED chips and reducing maintenance frequency. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an AI-based adaptive brightness adjustment method for LED beads. Figure 2 This is a schematic diagram of the tensor generation process for illumination fit evaluation. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1 This is a flowchart illustrating the AI-based adaptive brightness adjustment method for LED beads according to an embodiment of the present invention. Figure 1 As shown, the method includes: Spatiotemporal decoupling analysis was performed on the multidimensional spectral distribution data within the lighting space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis was conducted on the reflectance spectral characteristic data to obtain a color fidelity feature sequence. The illumination fluctuation characteristics and the color fidelity characteristic sequences are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms to obtain an illumination fit evaluation tensor that characterizes the object's color reproduction quality under the current illumination environment. Based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial position index in the lighting adaptability evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated. The theoretical contribution matrix of each LED to each point in space is calculated based on the spatial position and beam angle parameters of each LED. The initial normalized driving intensity vector of each LED is set, and a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED is introduced. The product of the theoretical contribution matrix and the initial normalized driving intensity vector is iteratively solved so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

[0023] In one optional implementation, spatiotemporal decoupling analysis is performed on the multidimensional spectral distribution data within the illumination space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis is then performed on the reflectance spectral characteristic data to obtain a color fidelity feature sequence, including: The multidimensional spectral distribution data, which includes spatial distribution information and temporal evolution information of different wavelength components within the visible light band, is collected within the illumination space using an array of photoelectric sensors. The time evolution information is subjected to multi-level wavelet decomposition. During the decomposition process, wavelet coefficients at each scale are extracted and cross-correlation analysis is performed to identify cross-scale coupled illumination fluctuation patterns. Spatial coherence analysis is performed on the spatial distribution information to obtain the synchronization and phase difference of illumination fluctuations between different sampling locations as spatial coupling features. The spatial coupling features are then fused with the cross-scale coupled illumination fluctuation patterns to obtain illumination fluctuation features at different scales that include spatiotemporal coupling relationship identifiers. The reflectance spectral characteristic data of the object surface in the illumination space under the current illumination conditions are collected by a distributed spectrometer. The reflectance spectral characteristic data are converted into the chromaticity space to obtain a measured chromaticity coordinate sequence. The color difference value sequence between the measured chromaticity coordinate sequence and the reference chromaticity coordinate sequence of the object surface under the standard illuminator is calculated. Each color difference value in the color difference value sequence is normalized and compared with the human eye color perception threshold to obtain the color fidelity characteristic sequence that characterizes the accuracy of color reproduction of the object surface.

[0024] Within the illumination space, multidimensional spectral distribution data is acquired using a photoelectric sensor array. This array consists of multiple photoelectric detection units with broadband spectral response capabilities, arranged in a pre-defined grid topology to cover the main functional areas of the illumination space. Each detection unit independently acquires radiant intensity information for different wavelength components within the visible light band (380nm to 780nm) and records the evolution of that wavelength component over time at a fixed sampling frequency. The acquisition results from all detection units for each wavelength component are integrated to form multidimensional spectral distribution data that simultaneously contains spatial distribution and temporal evolution information. The spatial distribution information describes the spectral radiation state at different locations within the illumination space at the same time, while the temporal evolution information describes the trajectory of spectral radiation intensity at a specific location as a function of sampling time. These two types of information together form the data foundation for subsequent spatiotemporal decoupling analysis.

[0025] For temporal evolution information, a multi-level wavelet decomposition method is employed. Wavelet decomposition simultaneously unfolds the temporal evolution information across multiple frequency scales, obtaining wavelet coefficient sequences at each decomposition level. Coarse-scale wavelet coefficients reflect the slow drift trend of light intensity, such as low-frequency fluctuations caused by changes in natural lighting over time; fine-scale wavelet coefficients capture rapid transient components caused by factors such as light switchover and the movement of obstructions. Cross-correlation analysis is performed on the wavelet coefficient sequences at each scale, calculating the cross-correlation function between coefficient sequences at different scales. Light fluctuation patterns that occur synchronously or have fixed time delay relationships across multiple frequency scales are identified. These statistically correlated fluctuation patterns across multiple scales are defined as cross-scale coupled light fluctuation patterns. Cross-scale coupled light fluctuation patterns can distinguish between local disturbances dominated by a single frequency component and systemic light change events that respond across a wide frequency band, providing more discriminative feature representations for subsequent pulse coding processing.

[0026] The spectral radiance sequences from different sampling locations at the same time are paired, and the coherence function between each pair is calculated to extract two types of parameters: coherence amplitude and phase difference. Coherence amplitude characterizes the synchronicity of illumination fluctuations at two locations, while phase difference characterizes the time delay of the fluctuations during spatial propagation; together, they constitute the spatial coupling feature. This spatial coupling feature reflects the propagation direction and diffusion speed of illumination changes within the lighting space. For example, illumination fluctuations near windows often occur before those deeper within the room; the corresponding phase difference quantifies this propagation delay. The spatial coupling feature is then fused with cross-scale coupled illumination fluctuation patterns in a spatiotemporally aligned manner. A spatiotemporal coupling relationship identifier is added to each spatiotemporal sampling point, containing the fluctuation pattern category to which the sampling point belongs and its phase position in the spatial propagation path. Finally, illumination fluctuation features at different scales containing the spatiotemporal coupling relationship identifier are obtained.

[0027] The reflectance spectrum of objects within the illuminated space is collected using a distributed spectrometer. Each probe of the distributed spectrometer is pointed at a different area of ​​the surface being measured, recording the reflectance spectral characteristics of the object's surface under the current illumination conditions—that is, the distribution curve of the surface's reflectance to incident light of various wavelengths as a function of wavelength. The reflectance spectral characteristics are then convolved with the spectral power distribution of the current illumination source, and integrated using the tristimulus function of the human eye. The calculation results are then converted to the CIE Lab color space to obtain the value in terms of lightness components. Red and green product weight Yellow and blue product weight The sequence of measured chromaticity coordinates is represented.

[0028] After obtaining the measured chromaticity coordinate sequence, using a D65 standard illuminator as the reference illumination condition, the chromaticity coordinates of the same object surface under the standard illuminator were calculated to obtain the reference chromaticity coordinate sequence. The color difference value was then calculated for each corresponding sampling point in both the measured and reference chromaticity coordinate sequences, using the CIE 2000 color difference formula. The formula incorporates perceptual uniformity correction weights across three dimensions: lightness, chroma, and hue, ensuring that the calculated color difference values ​​align with the subjective perception of color differences by the human eye. The color difference values ​​at each sampling point are then... The color difference value sequence is formed, and each element in the sequence corresponds to the color reproduction deviation of a specific measured surface area at a specific time.

[0029] Each color difference value in the color difference value sequence is compared with the normalized color perception threshold of the human eye. The color perception threshold of the human eye is usually taken as... As a reference boundary for the minimum perceptible color difference, when the color difference value is below this threshold, the human eye cannot perceive the color deviation; when the color difference value exceeds this threshold, the degree of deviation increases with the increase of the value. Normalization divides each color difference value by the perception threshold to obtain the relative deviation in units of the perception threshold. ,Right now ,in The threshold for color perception in the human eye. Regarding the relative deviation... A nonlinear mapping is performed, compressing the color reproduction to a range of 0 to 1. A mapping value close to 0 indicates that the color reproduction quality is close to that under standard illumination conditions, while a mapping value close to 1 indicates that the color reproduction deviation exceeds the acceptable range for the human eye. After the above normalization and comparison processing, the mapping results of each sampling point are arranged according to time and spatial location, forming a color fidelity feature sequence characterizing the accuracy of color reproduction on the object surface. This sequence describes the color reproduction quality of object surfaces of different materials and colors under the current illumination conditions using a unified perceptual dimension, providing standardized input data for subsequent joint analysis with illumination fluctuation characteristics.

[0030] The numerical distribution at different locations in the color fidelity feature sequence reveals the spatial non-uniformity of color reproduction quality within the lighting space. Areas directly beneath the luminaires typically exhibit lower relative deviations due to higher illuminance; conversely, areas at the edge of illumination or in strong shadows often show a significant decrease in color reproduction quality. Combining this spatial distribution information with the illumination fluctuation characteristics obtained from spatiotemporal decoupling analysis allows for the establishment of a correspondence between dynamic changes in the lighting environment and the real-time response of color reproduction quality, providing spatiotemporally consistent feature inputs for subsequent pulse code conversion and synaptic delay chain modeling.

[0031] In one optional implementation, the illumination fluctuation characteristics and the color fidelity characteristic sequence are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms. This yields an illumination fit evaluation tensor characterizing the object's color reproduction quality under the current illumination environment, comprising: The spatiotemporal coupling relationship identifier in the illumination fluctuation feature is mapped to the pulse emission spatiotemporal code, and the color fidelity feature sequence is mapped to the pulse emission frequency sequence; A synaptic delay chain containing multiple parallel delay paths is constructed. Each delay path corresponds to a time scale in the illumination fluctuation characteristics at different scales. The delay time parameters of each delay path are set according to the pulse emission spatiotemporal coding, so that the pulse signal in the pulse emission spatiotemporal coding generates a time delay effect corresponding to the illumination change rate when it propagates in the synaptic delay chain. A lateral inhibition mechanism is constructed, in which the inhibitory neurons dynamically adjust the lateral inhibition intensity according to the pulse firing frequency sequence. The lateral inhibition intensity is inversely proportional to the pulse frequency in the pulse firing frequency sequence, thereby realizing the encoding expression of the color response rate of the object surface. The output pulse sequence of the synaptic delay chain is time-aligned with the output pulse sequence of the lateral inhibition mechanism and the temporal correlation is calculated. The temporal correlation characterizes the temporal coupling relationship between the rate of change of illumination and the color response rate of the object surface. The illumination fit evaluation tensor is generated based on the temporal coupling relationship, and the illumination fit evaluation tensor includes the spatial location index in the pulse firing spatiotemporal coding as a spatial location index.

[0032] like Figure 2 As shown, the method includes: After acquiring the illumination fluctuation characteristics and color fidelity characteristic sequences, these two types of features need to be converted into a pulse coding form suitable for spiking neural network processing. For the illumination fluctuation characteristics, the spatiotemporal coupling indicators reflect the patterns of illumination intensity changes at different locations and times within the illumination space. When mapping these spatiotemporal coupling indicators to pulse firing spatiotemporal coding, the joint index of spatial coordinates and timestamps is used as the activation marker for spiking neurons. The amplitude of the abrupt change in illumination intensity at a certain spatial location determines the firing time of the corresponding neuron; the larger the amplitude, the earlier the firing time, thus encoding the intensity information of illumination changes into the pulse firing sequence. For the color fidelity characteristic sequence, the color reproduction quality index at each time point is mapped to the pulse firing frequency of the corresponding neuron; the higher the color reproduction quality, the higher the firing frequency, thus forming a pulse firing frequency sequence. This encoding method allows both illumination spatial information and color response information to enter the subsequent neural network processing flow in the form of pulse timing, providing a unified signal expression basis for subsequent temporal coupling analysis.

[0033] The synaptic delay chain contains multiple parallel delay paths, each corresponding to a specific time scale. These time scales correspond one-to-one with the multiple scales used in the spatiotemporal decoupling analysis stage to extract illumination fluctuation features. Let the number of delay paths be... , No. The delay time parameter of each delay path is ,in . The value is determined based on the first pulse firing time-space coding. Rate of change of light on a time scale Configure the settings: the faster the rate of change in illumination, the shorter the corresponding delay time; the slower the rate of change in illumination, the longer the corresponding delay time. Specifically, and The mapping relationship between them satisfies a monotonically decreasing relationship, causing rapid changes in illumination to produce short delays and slow changes in illumination to produce long delays. This naturally reproduces the time delay effect that matches the actual rhythm of illumination changes during the propagation of the pulse signal in the synaptic delay chain. After the pulse signal in the pulse firing spatiotemporal coding enters the synaptic delay chain, it undergoes... The output pulse sequences of each pathway together constitute the output of the synaptic delay chain after a time delay, carrying temporal information about the rate of change of illumination at different time scales.

[0034] Lateral inhibition mechanisms are constructed to encode the color response rate of object surfaces. In lateral inhibition mechanisms, inhibitory neurons receive a sequence of pulse firing frequencies as input, and then adjust the frequency based on the current pulse frequency. Dynamic adjustment of side inhibition intensity The intensity of lateral suppression is inversely proportional to the pulse frequency; that is, when color fidelity is high, the pulse frequency... When it is large, the lateral inhibition strength Smaller pulse frequencies allow more excitatory signals to pass through; when color fidelity is low and pulse frequency is low... When smaller, lateral inhibition strength Larger lateral inhibition levels exert a stronger inhibitory effect on the activation of surrounding neurons. This inverse regulation mechanism is a functional extension of the lateral inhibition principle in the biological visual system, allowing the speed of color response to be effectively distinguished by the intensity of lateral inhibition. Regions with faster color response rates correspond to higher lateral inhibition levels. Its lateral inhibition strength The lower the color response rate, the higher the effective pulse density in the output pulse sequence; conversely, the lower the color response rate, the lower the effective pulse density. Therefore, the output pulse sequence of the side suppression mechanism carries temporal coding information about the object's surface color response rate.

[0035] Timing alignment is achieved by unifying the time base of two pulse sequences to the same reference time, ensuring that pulse signals from different processing paths are comparable on the time axis. After timing alignment is completed, the temporal correlation between the two pulse sequences is calculated. ,in This is the time offset. Reflects the time offset Under certain conditions, the degree of co-firing between the output of the synaptic delay chain and the output of the lateral inhibition mechanism: when the two pulse sequences are at a certain... During altitude synchronization, A large value indicates a strong temporal coupling between the rate of illumination change and the color response rate on this time offset scale; when the two sequences are out of sync... A smaller value indicates weaker coupling between the two at that scale. This is achieved by traversing different... The value is used to obtain the complete temporal correlation function, thus comprehensively characterizing the temporal coupling relationship between the rate of illumination change and the color response rate of the object surface. This temporal coupling relationship includes both coupling strength information at different time scales and retains the corresponding spatial location information, providing a complete input basis for the subsequent generation of the illumination fit evaluation tensor.

[0036] When generating the illumination fit evaluation tensor based on the temporal coupling relationship, the spatial location index carried in the spatiotemporal encoding of the pulse firing is used as the spatial dimension index of the tensor, with different time scales. The corresponding temporal coupling strength is used as the feature dimension value of the tensor. Each spatial location element of the illumination adaptability evaluation tensor integrates the coupling degree between the illumination change rate and the color response rate at that location across multiple time scales. A higher coupling degree indicates a better match between the illumination conditions at that location and the object's color reproduction requirements, i.e., a higher illumination adaptability. A lower coupling degree indicates a significant deviation in the color reproduction quality at that location under the current illumination conditions, requiring compensation through subsequent driving intensity adjustments. The spatial location index in the illumination adaptability evaluation tensor is directly derived from the spatial coordinate markers established during the pulse firing spatiotemporal coding stage, ensuring a precise correspondence between each element in the tensor and the actual physical location in the illumination space, providing a reliable spatial positioning basis for the subsequent generation of the spatial illuminance demand distribution matrix. The entire pulse coding and temporal coupling analysis process achieves efficient expression of the dynamic relationship between the illumination environment and color response through neuromorphic computation, balancing the dual requirements of temporal resolution and spatial positioning accuracy.

[0037] In one optional implementation, based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial location index in the lighting fit evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated, including: Based on the spatial coordinate information and surface normal vector information in the three-dimensional geometric structure data, the reflectivity parameters and diffuse reflection direction distribution parameters of each surface are determined. Starting from the installation position of the LED beads, a tracking light is emitted to each surface. At each intersection point, a reflected light is generated according to the reflectivity parameter and the diffuse reflection direction distribution parameter. The propagation path of the light is iteratively tracked multiple times. The illumination contribution corresponding to each light propagation path is calculated and accumulated to obtain the total illumination contribution path of each surface. The multiple reflection attenuation coefficient is calculated based on the number of reflections in the light propagation path. The multiple reflection attenuation coefficient is equal to the product of the reflectivity parameters of each reflection in the light propagation path. Spatial matching is performed between the spatial location index in the illumination adaptability evaluation tensor and the spatial coordinate information to determine the surface position corresponding to each spatial location index. The illuminance requirement is calculated based on the total light contribution path at the surface location and the multiple reflection attenuation coefficient. The illuminance requirement is equal to the product of the total light contribution path and the reciprocal of the multiple reflection attenuation coefficient. The position priority weight is calculated based on the fitness value corresponding to each spatial position index in the illumination fitness evaluation tensor. The position priority weight is proportional to the fitness value. The illumination demand value and the position priority weight are combined to generate the spatial illumination demand distribution matrix.

[0038] After acquiring the 3D geometric data of the lighting space, it is necessary to extract spatial coordinate information and surface normal vector information. The 3D geometric data is typically imported from a 3D scanner, LiDAR, or Building Information Modeling (BIM), and includes the vertex coordinates, triangular topological relationships, and unit normal vectors of each facet of the walls, floors, ceilings, and furniture surfaces within the space. For each discretized surface facet, the corresponding reflectivity parameter is queried from its material property database. This parameter describes the proportion of incident light energy retained by the surface and ranges from 0 to 1 as a real number. Simultaneously, the diffuse reflection direction distribution parameter is determined based on the micro-roughness information of the surface material. Smooth surfaces tend towards specular reflection, while rough surfaces tend towards Lambertian diffuse reflection. In real-world scenarios, most building materials fall between these two extremes, and a hybrid bidirectional reflectance distribution function (BRDF) is used to describe this, where the sum of the diffuse reflection component weights and the specular reflection component weights is normalized to 1.

[0039] Starting from the installation position of the LED beads, tracking rays are emitted to various surfaces within the illumination space. The ray tracing employs a Monte Carlo path tracing method. Starting from the spatial coordinates of each LED bead, several initial directions are uniformly sampled within the solid angle range defined by its beam angle parameters, generating a corresponding number of tracking rays. Each ray propagates in space, detecting its intersection with each surface patch to determine the nearest intersection point. At the intersection point, based on the reflectivity and diffuse reflection direction distribution parameters of the surface patch, the outgoing direction is probabilistically sampled according to the BRDF model to generate the next reflected ray. This process continues to track the intersection points of the reflected ray with other surfaces, iterating until the number of reflections reaches a preset maximum tracking depth or the ray energy decays below a threshold. For each complete ray propagation path, its illuminance contribution is calculated. The contribution is determined by the initial luminous flux and the energy loss introduced by each reflection along the path. The contributions of all ray paths on the same surface patch are summed to obtain the total illuminance contribution path for that surface, which physically represents the estimated cumulative irradiance received by that surface under the current LED bead configuration.

[0040] The calculation of the multiple reflection attenuation coefficient directly depends on the reflectivity parameters at each reflection point along the light propagation path. Suppose a light path passes through... Second reflection, the first The secondary reflection occurs when the reflectivity parameter is... On the surface, the attenuation coefficient of multiple reflections corresponding to this path. Defined as the product of all reflectivity parameters, i.e. When the path contains only direct light ( )hour, A value of 1 indicates no reflection attenuation. As the number of reflections increases, The value decreases monotonically, reflecting the physical law that light energy is gradually absorbed by the surface during multiple reflections. In actual calculations, the attenuation coefficients of multiple light paths on the same surface are calculated separately, and then weighted and averaged according to the contribution of each path to obtain the comprehensive multiple reflection attenuation coefficient of the surface, which is used for the back-calculation of subsequent illuminance requirements.

[0041] The illumination fit evaluation tensor contains spatial location indices, each corresponding to the 3D coordinates of a discrete sampling point in the illumination space. These spatial location indices are spatially matched with surface coordinates in the 3D geometry data. Specifically, for each location index, the nearest Euclidean distance facet is searched within the set of surface faces, and the location index is assigned to the surface represented by that facet. For sampling points located within space (non-surface locations), spatial interpolation is used to map them to the nearest surface, or voxelization is employed for region merging. After spatial matching, each spatial location index is associated with a specific surface location, allowing subsequent calculations to utilize the total illumination contribution path and multiple reflection attenuation coefficients of that surface.

[0042] For a given surface location, let the estimated cumulative irradiance corresponding to its total illumination contribution path be valued as follows: The combined attenuation coefficient for multiple reflections is Then the illuminance requirement value at that surface location Defined as the product of the total illumination contribution path and the reciprocal of the multiple reflection attenuation coefficient, i.e. The physical meaning of this formula is that, due to energy loss during multiple reflections, a higher luminous flux output is needed at the light source to ensure sufficient illumination on the target surface. The reciprocal of the attenuation coefficient during multiple reflections is the compensation factor. When the surface reflectivity is low ( When the value is small, the compensation factor is large, and the corresponding illuminance requirement value is also increased, which is consistent with the physical intuition that dark material surfaces need stronger lighting in real-world scenarios.

[0043] After obtaining the illuminance requirements for each surface location, it is necessary to calculate the location priority weight by combining the fitness values ​​corresponding to the spatial location indices in the illumination fitness evaluation tensor. The fitness value reflects the color reproduction quality at that location under the current lighting environment. The lower the fitness value, the worse the color fidelity at that location, and the more urgent the need for improved lighting quality; therefore, a higher priority weight is assigned. Location Priority Weight Fit value Proportional to each other, it can be determined by linear normalization, that is, the fitness value of each position is normalized to the range of 0 to 1 in the global range and then directly used as the weight, so as to ensure the global consistency and comparability of weight allocation.

[0044] The illuminance demand values ​​are combined with location priority weights to generate a spatial illuminance demand distribution matrix. This matrix uses spatial location indices as row indices and weighted illuminance demand values ​​as matrix elements. Illuminance demand value With position priority weight The product is given, that is The spatial illuminance demand distribution matrix comprehensively describes the differentiated illuminance requirements at various locations within the lighting space. It includes both physical attenuation compensation information from multiple reflection paths and priority information assigned by color reproduction quality assessment, providing precise and physically meaningful target constraints for subsequent optimization of LED drive intensity. Locations with higher weights in the matrix will receive more illuminance resource allocation in subsequent iterative optimization processes, thereby achieving a directional tilt of lighting resources towards areas with weak color reproduction quality and improving the overall lighting adaptation level of the space.

[0045] In one optional implementation, the theoretical contribution matrix of each LED to each point in space is calculated based on the spatial position and beam angle parameters of each LED, an initial normalized driving intensity vector for each LED is set, and a performance degradation compensation mechanism based on the historical operating time and junction temperature accumulation of the LEDs is introduced, including: The beam angle parameters of each LED bead include the beam divergence angle and the spatial distribution characteristics of light intensity; Calculate the distance vector from each LED to each spatial point in the spatial illuminance demand distribution matrix based on the spatial position coordinates of each LED. Determine whether each spatial point is within the effective irradiation range of each LED based on the distance vector and the beam divergence angle. For spatial points within the effective irradiation range, calculate the theoretical illuminance contribution value of each LED to each spatial point based on the length of the distance vector and the light intensity spatial distribution characteristics, and organize it into a theoretical contribution matrix. Set the initial normalized driving intensity vector for each LED; The historical operating time data and junction temperature accumulation data of each LED bead are obtained. The operating time attenuation factor is calculated based on the historical operating time data. The operating time attenuation factor is inversely proportional to the historical operating time data. The temperature attenuation factor is calculated based on the junction temperature accumulation data. The temperature attenuation factor is inversely proportional to the junction temperature accumulation data. The comprehensive performance attenuation coefficient is obtained by multiplying the operating time attenuation factor and the temperature attenuation factor.

[0046] The beam angle parameters of each LED bead are the fundamental inputs for calculating the theoretical contribution matrix. These beam angle parameters comprise two core dimensions: beam divergence angle and spatial intensity distribution characteristics. The beam divergence angle determines the effective illumination cone range of a single LED bead, typically given as a half-angle, i.e., the angle from the LED bead's optical axis to where the light intensity drops to 50% of its peak value. The spatial intensity distribution characteristics describe how light intensity varies with the off-axis angle within the effective illumination cone. Common forms include Lambertian distribution (cosine distribution) and non-Lambertian distribution (such as batwing-shaped, focused, etc.). Different types of LED beads exhibit significantly different illumination effects in actual installation scenarios; therefore, it is necessary to model the spatial intensity distribution characteristics of each LED bead individually, rather than using a uniform approximation model.

[0047] Based on the spatial coordinates of each LED bead and the coordinates of each spatial point in the spatial illuminance demand distribution matrix, the distance vector from each LED bead to each spatial point is calculated. This distance vector contains not only the Euclidean distance between the two points but also directional information, i.e., the unit direction vector pointing from the LED bead to the spatial point. Taking the inner product of this direction vector and the LED bead's optical axis direction vector yields the off-axis angle of the spatial point relative to the LED bead's optical axis. .Will With the beam divergence angle If a comparison is made, If so, then the spatial point is determined to be within the effective illumination range of the corresponding LED; if If the LED is not selected, its theoretical contribution to the illumination intensity of that spatial point is set to zero and it is not included in subsequent calculations. This judgment process is performed on each combination of LEDs and spatial points to form a sparse and effective coverage mapping, providing structural constraints for subsequent matrix construction.

[0048] For the combination of lamp beads and spatial points determined to be within the effective illumination range, based on the length of the distance vector... With off-axis angle The spatial distribution characteristics of the light intensity of the LED bead are analyzed to calculate the theoretical contribution value of the illumination intensity. Taking a Lambertian distribution as an example, the change in light intensity with the off-axis angle satisfies... ,in This represents the peak luminous intensity along the optical axis of the LED. The order is Lambertian, determined by the LED chip's packaging structure. The irradiance reaching the spatial point also needs to consider the inverse squared attenuation over distance, i.e. ,in This represents the theoretical contribution of a single LED chip to the illuminance at a given point in space under a unit driving intensity. This represents the distance from the LED to this spatial point. This applies to all LEDs and all spatial points. The values ​​are arranged according to the spatial points of the row and the LED beads of the column, thus forming the theoretical contribution matrix. The matrix contains a large number of zero elements (corresponding to combinations outside the effective irradiation range), exhibiting an overall sparse structure, which is beneficial for optimizing the computational efficiency of subsequent iterative solutions.

[0049] Initial normalized driving intensity vector of each LED bead The initialization strategy employs uniform initialization, assigning all elements in the vector a pre-defined initial ratio value. This value is typically estimated based on the ratio of the total illuminance requirement to the rated power of the LEDs, ensuring that the overall illuminance level in the initial state falls within a reasonable range and preventing numerical divergence or slow convergence during the iteration process. In scenarios with prior lighting experience, adjustments can also be made based on historical adjustment data. Differential initialization is performed to match the initial driving intensity of the LEDs in different regions with the historical illuminance requirements of that region, thereby shortening the number of steps required for iterative convergence.

[0050] Obtaining historical operating time data and cumulative junction temperature data for each LED chip is a prerequisite for constructing a performance degradation compensation mechanism. Historical operating time data records the cumulative lighting time of each chip since it was put into use, typically stored in hours in the non-volatile memory of the lighting controller, and is automatically updated each time the lights are turned off. Cumulative junction temperature data reflects the accumulated thermal stress experienced by the chip during its historical operation. It is obtained by integrating or weighted summing the historical sampling data from the chip's junction temperature sensor, and can capture the differences in thermal damage under different operating conditions.

[0051] Calculate the duration decay factor based on historical working duration data. The duration decay factor is inversely proportional to historical working duration data, specifically expressed as follows: ,in The cumulative working time of the LED beads throughout history, The decay rate coefficient is obtained by fitting the light decay characteristic curve of the LED model. This formula ensures... exist When the value is zero, it takes the value of 1 (i.e., the new LED has no attenuation). The increase followed by a monotonically decreasing trend aligns with the physical law that the luminous flux of LED beads gradually decreases over time.

[0052] Calculate the temperature decay factor based on the cumulative junction temperature data. The temperature decay factor is inversely proportional to the cumulative junction temperature data, expressed as... ,in This refers to the cumulative junction temperature of the LED chip. This represents the temperature decay rate coefficient. Junction temperature accumulation. The calculation method involves integrating the portion of the historical junction temperature sampling sequence that exceeds the rated junction temperature reference value over time. This means that only the thermal stress contribution under overheating conditions is statistically analyzed, thus avoiding interference from sampling data within the normal operating temperature range on the attenuation assessment. The value is determined by the reliability test data of the LED chips. Different packaging processes and phosphor formulations correspond to different values ​​for LED chips. value.

[0053] Duration decay factor With temperature decay factor Multiplying these together yields the overall performance degradation coefficient for each LED chip. ,Right now Overall performance degradation coefficient The physical meaning is: the estimated ratio of the actual luminous flux that the LED can output under current operating conditions to its rated initial luminous flux. For LEDs with severe attenuation, A smaller value means that the actual light output is lower than the theoretical expectation under the same driving intensity. Therefore, the theoretical contribution matrix needs to be adjusted in subsequent iterative solutions. Multiply the elements of the corresponding column in the middle The system is modified to ensure that the matrix elements accurately reflect the actual light contribution capability of the LEDs, thereby driving iterative optimization to provide a higher normalized drive intensity compensation value to compensate for insufficient illuminance caused by performance degradation. This mechanism ensures that even in real-world scenarios where LEDs are aging or experiencing uneven thermal damage, the spatial illuminance distribution still meets the preset uniformity and adaptability requirements, extending the effective service life of the entire lighting system.

[0054] In one optional implementation, iteratively solving the product of the theoretical contribution matrix and the initial normalized driving intensity vector, such that the product approximates the spatial illuminance demand distribution matrix, includes: The initial illuminance distribution prediction value is obtained by performing matrix multiplication between the theoretical contribution matrix and the initial normalized driving intensity vector, and the illuminance difference matrix between the initial illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​calculated. The gradient direction is calculated on the initial normalized driving intensity vector based on the illuminance difference matrix. The gradient direction points in the direction that reduces the total error of the illuminance difference matrix. The initial normalized driving intensity vector is updated along the gradient direction to obtain an intermediate normalized driving intensity vector. Elements in the intermediate normalized driving intensity vector that are outside the normalized value range are subjected to projection constraints so that all elements remain within the normalized value range to obtain the constrained normalized driving intensity vector.

[0055] After obtaining the theoretical contribution matrix and the initial normalized driving intensity vector, an iterative optimization process is needed to make the actual output illuminance distribution of each LED chip approximate the spatial illuminance demand distribution matrix as closely as possible. The starting point for the iterative solution is to obtain the theoretical contribution matrix... With the initial normalized driving strength vector Perform matrix multiplication and record the result as the initial illuminance distribution prediction value. ,Right now This predicted value reflects the theoretical illuminance estimate that can be obtained at each sampling point in the lighting space under the current drive intensity configuration. Since the initial drive intensity vector is usually based on uniform or empirical initialization, there is often a large deviation between the corresponding predicted illuminance distribution and the actual illuminance requirement of the space. Therefore, it is necessary to further calculate the difference between the two and use this as the basis for driving iterative updates.

[0056] Illuminance difference matrix Defined as the matrix of predicted initial illuminance distribution and spatial illuminance demand distribution. The element-wise difference between them, i.e. In this difference matrix, positive values ​​indicate that the predicted illuminance at the corresponding spatial location is higher than the required value (too bright), while negative values ​​indicate that the predicted illuminance is lower than the required value (underbright). The sum of squares of the elements in the illuminance difference matrix is ​​used as a measure of the total error to construct the objective function. Its expression is ,in Let Frobenius norm be the norm of the matrix. This is achieved by considering the objective function with respect to the driving intensity vector. Taking the partial derivative gives the gradient direction. Its calculation formula is ,in This is the transpose of the theoretical contribution matrix. The gradient direction points in the direction of the fastest increase in the objective function value; therefore, updating the driving intensity vector along the opposite direction of the gradient can effectively reduce the total error of illumination difference.

[0057] In the In the next iteration, the step size is increased along the opposite direction of the gradient. For the current driving strength vector The intermediate normalized driving strength vector is then updated. Its update rules are as follows Step length The choice of step size has a significant impact on the convergence speed and stability of the iteration. Too large a step size can lead to iteration oscillations or even divergence, while too small a step size will result in slow convergence. Therefore, an adaptive step size search strategy based on the Armijo criterion can be adopted: in each iteration, the step size is adjusted from the initial step size... Departure, according to the reduction factor (The value range is usually from 0.5 to 0.8) Gradually reduce the step size until the sufficient descent condition is met, that is, the actual decrease in the objective function value is not less than a certain proportion of the theoretical linear approximation decrease. This strategy can accelerate the iteration speed while ensuring convergence stability.

[0058] Because the driving intensity of LED chips is physically constrained, the range of normalized driving intensity is: Where 0 corresponds to the completely off state and 1 corresponds to the full-power drive state. An intermediate normalized drive intensity vector is obtained during each gradient update. Next, projection constraints need to be applied to elements that exceed the range of values. Specifically, for... Each element Perform a truncation projection operation to map it to Within the interval, the driving intensity vector is normalized after obtaining the constraints. The projection rule is ,in This refers to the index number of the LED. This projection operation ensures that the driving intensity of all LEDs is within the legal physical operating range in any iteration step, avoiding negative driving or overload driving.

[0059] The projection constraint operation and the gradient update step together constitute the iterative framework of projective gradient descent. In each iteration, the predicted illuminance distribution and illuminance difference matrix are recalculated based on the current driving intensity vector, the gradient direction is updated, a step size search is performed, the gradient update of the driving intensity vector is completed, and projection constraints are applied to obtain the constraint-normalized driving intensity vector for the next iteration. Simultaneously, the theoretical contribution matrix needs to be dynamically adjusted according to the performance degradation compensation mechanism during the iteration process. The current overall performance attenuation coefficient of each LED bead. The contribution values ​​of the corresponding columns are scaled and corrected so that the iterative optimization is always based on the contribution matrix that reflects the actual performance state of the LEDs, thereby ensuring that the final driving intensity vector can truly compensate for the performance degradation differences of each LED.

[0060] The criteria for iterative convergence can be set from two dimensions: one is the magnitude of the change in the driving intensity vector between two adjacent iterations. Less than the preset convergence threshold Second, the current objective function value. The decrease compared to the previous iteration was lower than the preset convergence threshold. The iteration process terminates when any of the above conditions are met, and the constrained normalized driving strength vector obtained at this point is the final normalized driving strength vector. To prevent the iterative process from falling into a state of prolonged non-convergence due to the complexity of the objective function's terrain, a maximum upper limit on the number of iterations is also set. When the number of iterations reaches the upper limit, regardless of whether the convergence condition is met, the current driving strength vector is used as the final result output.

[0061] Final normalized driving intensity vector Each element corresponds to the normalized drive intensity value of an LED. Multiplying each element value by the rated drive current or rated PWM duty cycle of the corresponding LED generates a specific adjustment command. This command is sent to each LED driver chip via the control bus. The driver chip adjusts the actual operating current or PWM duty cycle of each LED according to the command, thereby achieving fine-grained adaptive adjustment of the illuminance distribution at various spatial locations within the lighting space. The entire iterative solution process uses the spatial illuminance demand distribution matrix as the optimization objective and the physically feasible drive intensity range as a constraint. The projected gradient descent iterative framework ensures the physical validity of the solution results and the precise matching of the illuminance distribution.

[0062] In one optional implementation, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism during the iteration process. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED bead, including: During the iteration process, the row vectors corresponding to each LED bead in the theoretical contribution matrix are scaled and adjusted according to the comprehensive performance attenuation coefficient. The scaling ratio of the scaling adjustment is equal to the comprehensive performance attenuation coefficient, thus obtaining the dynamically adjusted theoretical contribution matrix. The updated illuminance distribution prediction value is obtained by performing matrix multiplication between the dynamically adjusted theoretical contribution matrix and the constrained normalized driving intensity vector. The updated illuminance difference matrix between the updated illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​then calculated. Determine whether the total error of the updated illuminance difference matrix meets the convergence condition. The convergence condition is that the total error of the updated illuminance difference matrix is ​​less than a preset error threshold or the number of iterations reaches a preset maximum number of iterations. If the convergence condition is met, the constrained normalized driving intensity vector is determined as the final normalized driving intensity vector. If the convergence condition is not met, the constrained normalized driving intensity vector is used as the new initial normalized driving intensity vector and the gradient direction is recalculated. Based on the normalized drive intensity value of each LED in the final normalized drive intensity vector, an adjustment command is generated for the corresponding LED, and the adjustment command is applied to the drive circuit of the corresponding LED.

[0063] After performing gradient descent and projection constraint operations and obtaining the normalized driving intensity vector after constraints in each iteration step, the theoretical contribution matrix needs to be updated synchronously based on the current comprehensive performance attenuation coefficient of each LED chip. This ensures that the physical model upon which subsequent illuminance prediction relies always reflects the true light output capability of the LED chips. Specifically, the theoretical contribution matrix... Each row in the vector represents the contribution of a single LED bead to the illumination per unit driving intensity at each sampling point in space. When the performance of an LED bead degrades due to accumulated historical operating time or junction temperature, the optical contribution relationship described by that row vector is no longer accurate. Therefore, for the ... The row vector corresponding to each LED bead With its comprehensive performance attenuation coefficient As a scaling factor, the original row vector is multiplied by . The dynamically adjusted theoretical contribution matrix is ​​obtained. The corresponding row vector This operation scales the row vectors of all LEDs in the matrix according to their independent attenuation coefficients, thereby dynamically adjusting the theoretical contribution matrix. It can accurately characterize the actual illumination contribution of each LED in the current iteration step.

[0064] After completing the dynamic adjustment of the theoretical contribution matrix, With the current constraint-normalized driving strength vector Perform matrix multiplication to obtain the updated predicted illuminance distribution. ,Right now Each element of this predicted value represents the expected illuminance level at the corresponding sampling point in space under the current driving scheme and attenuation correction. Subsequently, Spatial Illuminance Demand Distribution Matrix Subtracting element by element yields the updated illuminance difference matrix. This difference matrix reflects the illuminance deviation of the current driving scheme at various spatial locations; positive values ​​indicate insufficient illuminance, and negative values ​​indicate excessive illuminance. Through analysis of... The sum of the squares of all elements in the matrix yields the total error metric for the current iteration step. This value is directly used for subsequent convergence judgment.

[0065] The convergence condition is determined based on two criteria: one is the total error in updating the illuminance difference matrix. Is it less than the preset error threshold? Second, the current iteration number. Has the preset maximum number of iterations been reached? If either of the two conditions is satisfied, the iteration is considered to have converged. When the current driving scheme can meet the spatial illumination requirements with sufficiently high accuracy, further iterations will no longer improve the results beyond the accuracy threshold required for practical applications, and stopping the iteration is reasonable. When the number of iterations reaches [a certain threshold], [the process continues]. Even if the total error is not lower than This also forces the termination of iterations to prevent excessive system response delays due to slow convergence speeds in complex lighting scenarios, which could negatively impact real-time adjustment performance. In actual engineering deployments, The value should be determined in conjunction with the spatial illuminance demand distribution matrix. The magnitude is normalized, and it can usually be set to... A fixed percentage of the sum of squares of all elements in the threshold is used to make the convergence threshold adaptively scale-independent.

[0066] When the convergence condition is met, the current constraint-normalized driving intensity vector is then applied. Determined as the final normalized driving strength vector If the convergence condition is not met, then... As the initial driving force vector for the next iteration, it is re-based on the dynamically adjusted theoretical contribution matrix. Calculate the objective function Regarding the gradient direction of the driving intensity vector Continue performing gradient descent and projection constraint operations. It is important to note that before each new iteration, the theoretical contribution matrix must be updated based on the latest overall performance decay coefficient of each LED, rather than using the dynamic adjustment results from the previous round. This ensures that the physical model remains consistent with the actual state of the LEDs throughout the entire iteration process.

[0067] After iterative convergence, the driving intensity vector is finally normalized. The Middle element Indicates the first The normalized drive intensity that each LED chip should output is constrained to a range between 0 and 1. This normalized value is then mapped to the actual control range supported by the corresponding LED chip driver circuit, such as in PWM duty cycle control. Multiplying this by the full-scale PWM count value of the drive circuit yields the corresponding duty cycle setting value; in analog current control mode, then... Multiply by the rated drive current of the LED bead to obtain the target drive current value. When generating the adjustment command, it must include both the LED bead address encoding information and the target control quantity value to ensure that the command can be accurately routed to the corresponding LED bead's drive circuit control register. The adjustment command is sent to each LED bead driver chip via the lighting control bus (such as DALI bus, I²C bus, or dedicated PWM signal line). After receiving the command, the driver chip smoothly transitions to the target drive quantity according to the set ramp rate, avoiding flickering perceptible to the human eye caused by sudden changes in drive intensity. The ramp rate setting should refer to the human eye's time perception characteristics of brightness changes. The transition time is usually set in the range of 50 milliseconds to 200 milliseconds and can be dynamically configured according to the urgency of scene switching. The entire process from iterative convergence to command application forms a closed loop, ensuring that the LED bead array, while taking into account performance degradation compensation, always meets the color reproduction and illuminance distribution requirements of the lighting space with the optimal drive scheme.

[0068] A second aspect of this invention provides an AI-based adaptive brightness adjustment system for LED beads, comprising: The spectral analysis unit is used to perform spatiotemporal decoupling analysis on multidimensional spectral distribution data in the lighting space, extract illumination fluctuation characteristics at different scales, perform colorimetric analysis on reflectance spectral characteristic data, and obtain color fidelity characteristic sequences. The temporal coupling unit is used to convert the light fluctuation characteristics and the color fidelity characteristic sequence into pulse code form, and establish the temporal coupling relationship between the light change rate and the object surface color response rate through synaptic delay chain and lateral inhibition mechanism to obtain the light fit evaluation tensor characterizing the color reproduction quality of the object under the current lighting environment. The path requirement unit is used to calculate the light contribution path and multiple reflection attenuation coefficient of each surface in the space based on the three-dimensional geometric structure data of the lighting space, and generate a spatial illuminance requirement distribution matrix by combining the spatial position index in the light fit evaluation tensor. The drive compensation unit is used to calculate the theoretical contribution matrix of each LED to each point in space based on the spatial position and beam angle parameters of each LED, set the initial normalized drive intensity vector of each LED, and introduce a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED. The iterative driving unit is used to iteratively solve the product of the theoretical contribution matrix and the initial normalized driving intensity vector, so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance decay compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

[0069] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0070] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0071] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An AI-based adaptive brightness adjustment method for LED beads, characterized in that, include: Spatiotemporal decoupling analysis was performed on the multidimensional spectral distribution data within the lighting space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis was conducted on the reflectance spectral characteristic data to obtain a color fidelity feature sequence. The illumination fluctuation characteristics and the color fidelity characteristic sequences are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms to obtain an illumination fit evaluation tensor that characterizes the object's color reproduction quality under the current illumination environment. Based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial position index in the lighting adaptability evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated. The theoretical contribution matrix of each LED to each point in space is calculated based on the spatial position and beam angle parameters of each LED. The initial normalized driving intensity vector of each LED is set, and a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED is introduced. The product of the theoretical contribution matrix and the initial normalized driving intensity vector is iteratively solved so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

2. The method according to claim 1, characterized in that, Spatiotemporal decoupling analysis was performed on the multidimensional spectral distribution data within the lighting space to extract illumination fluctuation characteristics at different scales. Colorimetric analysis was conducted on the reflectance spectral characteristic data to obtain the color fidelity feature sequence, including: The multidimensional spectral distribution data, which includes spatial distribution information and temporal evolution information of different wavelength components within the visible light band, is collected within the illumination space using an array of photoelectric sensors. The time evolution information is subjected to multi-level wavelet decomposition. During the decomposition process, wavelet coefficients at each scale are extracted and cross-correlation analysis is performed to identify cross-scale coupled illumination fluctuation patterns. Spatial coherence analysis is performed on the spatial distribution information to obtain the synchronization and phase difference of illumination fluctuations between different sampling locations as spatial coupling features. The spatial coupling features are then fused with the cross-scale coupled illumination fluctuation patterns to obtain illumination fluctuation features at different scales that include spatiotemporal coupling relationship identifiers. The reflectance spectral characteristic data of the object surface in the illumination space under the current illumination conditions are collected by a distributed spectrometer. The reflectance spectral characteristic data are converted into the chromaticity space to obtain a measured chromaticity coordinate sequence. The color difference value sequence between the measured chromaticity coordinate sequence and the reference chromaticity coordinate sequence of the object surface under the standard illuminator is calculated. Each color difference value in the color difference value sequence is normalized and compared with the human eye color perception threshold to obtain the color fidelity characteristic sequence that characterizes the accuracy of color reproduction of the object surface.

3. The method according to claim 2, characterized in that, The illumination fluctuation characteristics and the color fidelity characteristic sequences are converted into pulse-coded form. A temporal coupling relationship between the illumination change rate and the object surface color response rate is established through synaptic delay chains and lateral inhibition mechanisms. This yields an illumination fit evaluation tensor characterizing the object's color reproduction quality under the current illumination environment, including: The spatiotemporal coupling relationship identifier in the illumination fluctuation feature is mapped to the pulse emission spatiotemporal code, and the color fidelity feature sequence is mapped to the pulse emission frequency sequence; A synaptic delay chain containing multiple parallel delay paths is constructed. Each delay path corresponds to a time scale in the illumination fluctuation characteristics at different scales. The delay time parameters of each delay path are set according to the pulse emission spatiotemporal coding, so that the pulse signal in the pulse emission spatiotemporal coding generates a time delay effect corresponding to the illumination change rate when it propagates in the synaptic delay chain. A lateral inhibition mechanism is constructed, in which the inhibitory neurons dynamically adjust the lateral inhibition intensity according to the pulse firing frequency sequence. The lateral inhibition intensity is inversely proportional to the pulse frequency in the pulse firing frequency sequence, thereby realizing the encoding expression of the color response rate of the object surface. The output pulse sequence of the synaptic delay chain is time-aligned with the output pulse sequence of the lateral inhibition mechanism and the temporal correlation is calculated. The temporal correlation characterizes the temporal coupling relationship between the rate of change of illumination and the color response rate of the object surface. The illumination fit evaluation tensor is generated based on the temporal coupling relationship, and the illumination fit evaluation tensor includes the spatial location index in the pulse firing spatiotemporal coding as a spatial location index.

4. The method according to claim 1, characterized in that, Based on the three-dimensional geometric data of the lighting space, the light contribution path and multiple reflection attenuation coefficient of each surface in the space are calculated. Combined with the spatial location index in the lighting fit evaluation tensor, a spatial illuminance demand distribution matrix is ​​generated, including: Based on the spatial coordinate information and surface normal vector information in the three-dimensional geometric structure data, the reflectivity parameters and diffuse reflection direction distribution parameters of each surface are determined. Starting from the installation position of the LED beads, a tracking light is emitted to each surface. At each intersection point, a reflected light is generated according to the reflectivity parameter and the diffuse reflection direction distribution parameter. The propagation path of the light is iteratively tracked multiple times. The illumination contribution corresponding to each light propagation path is calculated and accumulated to obtain the total illumination contribution path of each surface. The multiple reflection attenuation coefficient is calculated based on the number of reflections in the light propagation path. The multiple reflection attenuation coefficient is equal to the product of the reflectivity parameters of each reflection in the light propagation path. Spatial matching is performed between the spatial location index in the illumination adaptability evaluation tensor and the spatial coordinate information to determine the surface position corresponding to each spatial location index. The illuminance requirement is calculated based on the total light contribution path at the surface location and the multiple reflection attenuation coefficient. The illuminance requirement is equal to the product of the total light contribution path and the reciprocal of the multiple reflection attenuation coefficient. The position priority weight is calculated based on the fitness value corresponding to each spatial position index in the illumination fitness evaluation tensor. The position priority weight is proportional to the fitness value. The illumination demand value and the position priority weight are combined to generate the spatial illumination demand distribution matrix.

5. The method according to claim 1, characterized in that, Based on the spatial position and beam angle parameters of each LED, the theoretical contribution matrix of each LED to each point in space is calculated. An initial normalized driving intensity vector for each LED is set. A performance degradation compensation mechanism based on the historical operating time and junction temperature accumulation of the LEDs is introduced, including: The beam angle parameters of each LED bead include the beam divergence angle and the spatial distribution characteristics of light intensity; Calculate the distance vector from each LED to each spatial point in the spatial illuminance demand distribution matrix based on the spatial position coordinates of each LED. Determine whether each spatial point is within the effective irradiation range of each LED based on the distance vector and the beam divergence angle. For spatial points within the effective irradiation range, calculate the theoretical illuminance contribution value of each LED to each spatial point based on the length of the distance vector and the light intensity spatial distribution characteristics, and organize it into a theoretical contribution matrix. Set the initial normalized driving intensity vector for each LED; The historical operating time data and junction temperature accumulation data of each LED bead are obtained. The operating time attenuation factor is calculated based on the historical operating time data. The operating time attenuation factor is inversely proportional to the historical operating time data. The temperature attenuation factor is calculated based on the junction temperature accumulation data. The temperature attenuation factor is inversely proportional to the junction temperature accumulation data. The comprehensive performance attenuation coefficient is obtained by multiplying the operating time attenuation factor and the temperature attenuation factor.

6. The method according to claim 5, characterized in that, Iteratively solving for the product of the theoretical contribution matrix and the initial normalized driving intensity vector, so that the product approximates the spatial illuminance demand distribution matrix, includes: The initial illuminance distribution prediction value is obtained by performing matrix multiplication between the theoretical contribution matrix and the initial normalized driving intensity vector, and the illuminance difference matrix between the initial illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​calculated. The gradient direction is calculated on the initial normalized driving intensity vector based on the illuminance difference matrix. The gradient direction points in the direction that reduces the total error of the illuminance difference matrix. The initial normalized driving intensity vector is updated along the gradient direction to obtain an intermediate normalized driving intensity vector. Elements in the intermediate normalized driving intensity vector that are outside the normalized value range are subjected to projection constraints so that all elements remain within the normalized value range to obtain the constrained normalized driving intensity vector.

7. The method according to claim 6, characterized in that, During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance degradation compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads, including: During the iteration process, the row vectors corresponding to each LED bead in the theoretical contribution matrix are scaled and adjusted according to the comprehensive performance attenuation coefficient. The scaling ratio of the scaling adjustment is equal to the comprehensive performance attenuation coefficient, thus obtaining the dynamically adjusted theoretical contribution matrix. The updated illuminance distribution prediction value is obtained by performing matrix multiplication between the dynamically adjusted theoretical contribution matrix and the constrained normalized driving intensity vector. The updated illuminance difference matrix between the updated illuminance distribution prediction value and the spatial illuminance demand distribution matrix is ​​then calculated. Determine whether the total error of the updated illuminance difference matrix meets the convergence condition. The convergence condition is that the total error of the updated illuminance difference matrix is ​​less than a preset error threshold or the number of iterations reaches a preset maximum number of iterations. If the convergence condition is met, the constrained normalized driving intensity vector is determined as the final normalized driving intensity vector. If the convergence condition is not met, the constrained normalized driving intensity vector is used as the new initial normalized driving intensity vector and the gradient direction is recalculated. Based on the normalized drive intensity value of each LED in the final normalized drive intensity vector, an adjustment command is generated for the corresponding LED, and the adjustment command is applied to the drive circuit of the corresponding LED.

8. An AI-based adaptive brightness adjustment system for LED beads, used to implement the method as described in any one of claims 1-7, characterized in that, include: The spectral analysis unit is used to perform spatiotemporal decoupling analysis on multidimensional spectral distribution data in the lighting space, extract illumination fluctuation characteristics at different scales, perform colorimetric analysis on reflectance spectral characteristic data, and obtain color fidelity characteristic sequences. The temporal coupling unit is used to convert the light fluctuation characteristics and the color fidelity characteristic sequence into pulse code form, and establish the temporal coupling relationship between the light change rate and the object surface color response rate through synaptic delay chain and lateral inhibition mechanism to obtain the light fit evaluation tensor characterizing the color reproduction quality of the object under the current lighting environment. The path requirement unit is used to calculate the light contribution path and multiple reflection attenuation coefficient of each surface in the space based on the three-dimensional geometric structure data of the lighting space, and generate a spatial illuminance requirement distribution matrix by combining the spatial position index in the light fit evaluation tensor. The drive compensation unit is used to calculate the theoretical contribution matrix of each LED to each point in space based on the spatial position and beam angle parameters of each LED, set the initial normalized drive intensity vector of each LED, and introduce a performance degradation compensation mechanism based on the historical working time and junction temperature accumulation of the LED. The iterative driving unit is used to iteratively solve the product of the theoretical contribution matrix and the initial normalized driving intensity vector, so that the product approximates the spatial illuminance demand distribution matrix. During the iteration process, the theoretical contribution matrix is ​​dynamically adjusted according to the performance decay compensation mechanism. After the iteration converges, the final normalized driving intensity vector is obtained and converted into an adjustment command to be applied to the corresponding LED beads.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images