LED lamp strip dynamic image quality intelligent optimization method and system based on deep learning driving

Abstract

The invention discloses an LED lamp strip dynamic image quality intelligent optimization method and system based on deep learning driving, and relates to the technical field of LED intelligent control. The LED lamp strip dynamic image quality intelligent optimization method and system based on deep learning driving comprises the steps of S1, preprocessing real-time LED working condition data, memory reference data and historical fluctuation abnormal data; s2, adaptively dividing the LED lamp strip into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering; s3, based on a time sequence derivative sensitivity theory, judging thermal drift instability and color temperature abnormity, and entering a sudden change criterion process; s4, executing a sudden change criterion process according to a trend sudden change composite criterion result, and entering an offset compensation process; and S5, based on a criterion compensation theory, executing an offset compensation process. The problems of image quality degradation and color difference accumulation after long-term operation due to the fact that an existing LED lamp strip image quality intelligent optimization system is difficult to achieve real-time self-adaptive compensation of color temperature drifting and color cast phenomena caused by chip junction temperature rise are solved.

Description

Technical Field

[0001] This invention relates to the field of LED intelligent control technology, specifically to a method and system for intelligent optimization of dynamic image quality of LED light strips based on deep learning. Background Technology

[0002] With the increasing demands for image quality in the lighting and display industries, intelligent optimization technology has become a crucial research direction in these fields. In recent years, LEDs, as a new generation of high-efficiency light sources, have been widely used in various intelligent lighting, display, and landscape systems due to their energy-saving, long lifespan, and fast response speed. As application demands continue to rise, the technology sector has conducted extensive research and optimization on the performance, reliability, and operational status monitoring of LED light strips. To further improve the color consistency, brightness stability, and visual comfort of LED applications, the industry is gradually introducing intelligent optimization technologies such as signal processing, data analysis, and adaptive control to achieve real-time adjustment and compensation of light source image quality. Intelligent optimization is providing continuous technical support for image quality management and automatic optimization.

[0003] For example, invention patent CN112541876A discloses a satellite image processing method, a network training method, and related devices and electronic equipment. The method includes: acquiring a satellite image to be processed and its overlapping extended satellite images; stitching them together to obtain a target satellite image; performing dehazing processing on the target satellite image based on a first neural network to obtain a dehazed satellite image; and further adjusting the image quality parameters of the satellite image through a second neural network to finally obtain a satellite image with optimized image quality. It also includes using a training image set, extracting features using a neural network, fusing target features, and reconstructing the image; continuously optimizing network parameters by comparing the differences between the output image and the target image, thereby achieving joint optimization of satellite image dehazing and image quality adjustment.

[0004] For example, the invention patent with publication number CN116188346A discloses a method and apparatus for enhancing the image quality of endoscopic images. The method includes: acquiring a complex image quality enhancement model, a lightweight image quality enhancement model, a set of original endoscopic images, and a set of low-resolution images; inputting the low-resolution image set into the two models respectively, outputting image sets with different image quality enhancements; generating a blurred endoscopic image set from the low-resolution image set using an edge-preserving filtering algorithm; selecting intermediate feature blocks at different depths in the two models and calculating the corresponding intermediate feature losses; comprehensively calculating the contrast loss and reconstruction loss between the image quality enhancement image sets; synthesizing the total loss based on the intermediate feature loss, contrast loss, and reconstruction loss; and then training the lightweight model using the complex model as the teacher network.

[0005] While existing intelligent image optimization and image quality enhancement methods can achieve functions such as defogging, noise reduction, and adaptive parameter adjustment for specific scenarios, they usually rely on single-task optimization, single-channel, or static strategies. They are difficult to detect and dynamically respond to the coupled changes of junction temperature and multiple physical parameters in a timely manner, and lack accurate identification and continuous optimization mechanisms for thermal color anomalies. During long-term operation, LED light strips cannot perform real-time, collaborative, and adaptive compensation for color temperature drift and color shift caused by the rise in chip junction temperature.

[0006] Therefore, in order to address the above problems, there is an urgent need for a method and system for intelligent optimization of dynamic image quality of LED light strips based on deep learning. Summary of the Invention

[0007] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a method and system for intelligent dynamic image quality optimization of LED light strips based on deep learning. This solves the problem that existing intelligent image quality optimization systems for LED light strips are unable to achieve real-time adaptive compensation for color temperature drift and color shift caused by chip junction temperature rise, leading to image quality degradation and color difference accumulation after long-term operation.

[0008] Technical solution To achieve the above objectives, this invention provides the following technical solution: a method and system for intelligent dynamic image quality optimization of LED light strips driven by deep learning, comprising: S1, collecting real-time LED operating condition data, acquiring memory benchmark data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory benchmark data, and historical fluctuation anomaly data; S2, adaptively dividing the LED light strip into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering; S3, based on the temporal derivative sensitivity theory, analyzing the dynamic characteristics of the LED light strip partitions in real time, identifying thermal drift instability and color temperature anomalies, and entering the mutation criterion process; S4, performing trend mutation composite criterion on the real-time LED operating condition data and memory benchmark data, executing the mutation criterion process according to the trend mutation composite criterion result, identifying and diverting chronic drift and sudden anomalies, and entering the offset compensation process; S5, analyzing the real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory, and executing the offset compensation process according to the analysis result of the criterion compensation theory.

[0009] Furthermore, real-time LED operating condition data is collected, and memory reference data and historical fluctuation anomaly data are obtained. The specific process for preprocessing the real-time LED operating condition data, memory reference data, and historical fluctuation anomaly data is as follows: Real-time LED operating condition data is collected, including: junction temperature timing signal, primary color temperature timing signal, raw illuminance data, primary peak wavelength sequence, real-time supply voltage, real-time loop current, and primary color luminous intensity sequence; memory reference data is obtained... The memory reference data includes: memory illuminance data, memory peak data, memory power data, memory color temperature data, memory voltage, memory current, and a sliding window for criterion features; historical fluctuation and anomaly data are acquired and a historical fluctuation and anomaly database is established, including: historical junction temperature fluctuation data, historical main color temperature drift data, historical main color deviation drift data, and historical anomaly data; noise suppression and smoothing are performed on the junction temperature time series signal, main color temperature time series signal, and raw illuminance data through median filtering and moving average algorithms to eliminate short-term pulse interference and high-frequency noise during the acquisition process; short-term breakpoints and anomaly missing values ​​are continuously filled in the memory reference data through linear interpolation and missing value filling algorithms; preliminary anomaly point identification and removal are performed on the historical fluctuation and anomaly data through batch statistics and box plot outlier removal algorithms, removing extreme outliers and retaining representative drift and anomaly trends; distribution standardization and linear normalization algorithms are used to standardize and normalize the real-time LED operating condition data, memory reference data, and historical fluctuation and anomaly data.

[0010] Furthermore, the specific process of adaptively dividing the LED light strip into different independent partitions according to spatial and physical attributes through multidimensional feature analysis and clustering is as follows: The LED light strip is partitioned according to spatial layout and operating characteristics. Combining the physical arrangement structure of the LED light strip, the grouping of driving circuits, and the thermo-optical response characteristics, the spatial distance is obtained by measuring the difference in actual physical coordinates. Adjacency is determined based on the physical connection and grouping relationship of LED units. Feature similarity is obtained by the Euclidean distance between the junction temperature time series signal, the main color temperature time series signal, and the original illuminance data. Dimensionality reduction processing is performed on the junction temperature time series signal, the main color temperature time series signal, and the original illuminance data based on principal component analysis to extract the main spatial distribution features. Then, the spectral clustering algorithm is used to combine spatial distance, adjacency, and feature similarity indexes to realize multidimensional feature segmentation of the LED light strip, dividing the LED light strip into multiple partitions. Each partition contains several LED units with continuous structure and similar physical attributes. Independent data acquisition channels and operating parameter records are established for each partition.

[0011] Furthermore, based on the time-series derivative sensitivity theory, the specific process for real-time analysis of the dynamic characteristics of LED strip zones is as follows: Obtain the junction temperature time-series signal and the main color temperature time-series signal; obtain the temperature change rate of the junction temperature time-series signal using a sliding window algorithm, and calculate the variance of the zone temperature change rate using the variance within the window; obtain the mean of the zone temperature change rate of the junction temperature time-series signal using both the sliding window algorithm and the mean algorithm; obtain the second derivative by performing a second-order difference operation on the main color temperature time-series signal within the sliding window, extract the maximum absolute value of the second derivative, and obtain the second-order extreme value of the zone color temperature offset; obtain the second-order extreme value of the zone color temperature offset using a difference algorithm on the main color temperature time-series signal. To obtain the color temperature shift sequence, the standard deviation of the color temperature shift for each partition is obtained using the sliding window standard deviation algorithm. The entropy of the color temperature shift for each partition is obtained using the Shannon entropy calculation algorithm. The ratio of the variance of the partition temperature change rate to the sum of the absolute values ​​of the mean of the partition temperature change rate is calculated to obtain the partition temperature dynamic variation coefficient term. The ratio of the second-order extreme value of the partition color temperature shift sequence to the sum of the partition color temperature shift standard deviations is calculated, and multiplied by the partition color temperature shift information entropy to obtain the partition color temperature dynamic complexity term. The larger value of the partition temperature dynamic variation coefficient term and the partition color temperature dynamic complexity term is taken to obtain the partition thermal color deviation instability value.

[0012] Furthermore, the specific process for identifying thermal drift instability and color temperature anomalies and entering the mutation criterion process is as follows: Real-time comparison of the zone's thermal color deviation instability value and the zone's thermal color deviation instability threshold. The zone's thermal color deviation instability threshold includes a primary instability threshold and a secondary instability threshold. When the zone's thermal color deviation instability value is less than the secondary instability threshold, it is determined to be in a safe zone, and the zone operates under normal current conditions without special intervention. When the zone's thermal color deviation instability value is greater than or equal to the secondary instability threshold but less than the primary instability threshold, it is determined to be in a warning zone. The LED drive current is reduced, the real-time LED operating condition data sampling frequency is increased, and the zone enters the mutation criterion process. When the zone's thermal color deviation instability value is greater than or equal to the primary instability threshold, it is determined to be in a risk zone. The zone's relay output and electronic switch are disconnected, the zone's heat sink device is activated, a zone red alarm signal is triggered, and a risk event is created and recorded in the risk event database, and a maintenance reminder is sent.

[0013] Furthermore, the specific process of applying a trend change composite criterion to real-time LED operating condition data and memory reference data is as follows: Acquire raw illuminance data, memory illuminance data, peak wavelength sequence, memory peak data, main color temperature time-series signal, memory color temperature data, real-time supply voltage, memory voltage, real-time loop current, and memory current; calculate the difference between the acquired raw illuminance data and the memory illuminance data using a sliding window algorithm to obtain the partitioned brightness offset; calculate the difference between the peak wavelength sequence and the memory peak data using a sliding window algorithm to obtain the peak wavelength offset; calculate the difference between the main color temperature time-series signal and the memory color temperature data to obtain the color temperature offset; calculate the difference between the partitioned real-time supply voltage and the memory voltage to obtain the voltage offset; calculate the difference between the partitioned real-time loop current and the memory current to obtain the current offset; multiply the real-time supply voltage and the real-time loop current to obtain the real-time power, and calculate the real-time power and the memory power data. The difference yields the power offset value; the product of the luminance offset value and the peak wavelength offset value is calculated, and the variance of this product within the criterion feature sliding window is statistically calculated. The absolute value of the mean of the criterion feature sliding window for the power offset value is added by one to obtain the power mean term. The variance result is divided by the power mean term, and the arithmetic square root is taken to obtain the luminance-wavelength joint fluctuation term; the product of the color temperature offset value and the sine of the voltage offset value is calculated, and the derivative of the mean of the criterion feature sliding window with respect to time is obtained. This derivative is multiplied by the trend change sensitivity factor, and the absolute value is taken to obtain the trend change response term; the difference between the maximum and minimum values ​​of the product of the peak wavelength offset value and the current offset value within the criterion feature sliding window is calculated, and the standard deviation of the luminance offset value within the criterion feature sliding window is added by one. The range result is divided by the standard deviation term, multiplied by the mutation response adjustment factor, and used as the base of the exponent. The exponential function value is taken to obtain the mutation range amplification term; the luminance-wavelength joint fluctuation term, the trend change response term, and the mutation range amplification term are added together to obtain the mutation criterion value.

[0014] Furthermore, based on the trend mutation composite criterion result, the mutation criterion process is executed to identify and triage chronic drift and sudden anomalies, and then proceeds to the offset compensation process. The specific process is as follows: real-time comparison of mutation criterion value and mutation criterion threshold, execution of the mutation criterion process; when the mutation criterion value is less than the mutation criterion threshold, it is determined to be a chronic drift region, with chip junction temperature rise; drift data and the entire process of partition drift are recorded, a chronic drift log is generated, and the offset compensation process is initiated; when the mutation criterion value is greater than or equal to the mutation criterion threshold, it is determined to be a sudden anomaly region; the LED drive circuit is immediately disconnected, partition safety lock is initiated, a partition red alarm signal is triggered, the sudden anomaly event is recorded to the risk event database, and a maintenance reminder is sent.

[0015] Furthermore, the specific process of analyzing real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory is as follows: Obtain the junction temperature time-series signal, main color temperature time-series signal, main peak wavelength sequence, main color intensity sequence, historical junction temperature fluctuation data, historical main color temperature drift data, historical main color deviation drift data, and historical anomaly data; obtain the main color deviation time-series signal from the main peak wavelength sequence and main color intensity sequence using the CIE 1976 uniform chromaticity space transformation algorithm; obtain the cross-correlation information entropy from the junction temperature time-series signal and main color temperature time-series signal using the sliding window cross-correlation analysis algorithm and the Shannon entropy calculation algorithm; obtain the junction temperature and main color temperature coupling characteristic quantity from the junction temperature time-series signal and main color temperature time-series signal using the sliding window cross-correlation analysis algorithm; obtain the junction temperature and main color deviation coupling characteristic quantity from the junction temperature time-series signal and main color deviation time-series signal using the sliding window cross-correlation analysis algorithm; calculate the power of the junction temperature response enhancement factor of the junction temperature change rate and the main color temperature change rate. The product of the thermal response enhancement factor raised to a power yields the thermal color response enhancement term; the power of the primary color deviation response suppression factor, calculated based on the primary color deviation change rate, plus one, yields the primary color deviation suppression term; the thermal response enhancement term is divided by the primary color deviation suppression term to obtain the coupling ratio term; the cross-correlation entropy of the junction temperature primary color temperature coupling characteristic is calculated to obtain the coupling information entropy term; the square of the junction temperature primary color deviation coupling characteristic is multiplied by the abnormal activation adjustment factor to calculate the exponential function, yielding the abnormal activation amplification term; the coupling ratio term, coupling information entropy term, and abnormal activation amplification term are multiplied together to obtain the offset compensation value.

[0016] Furthermore, the specific process of executing the offset compensation procedure based on the criterion compensation theory analysis results is as follows: The offset compensation value and the offset compensation threshold are compared in real time, and the offset compensation procedure is executed. When the offset compensation value is less than or equal to the offset compensation threshold, the sigmoid function is used to map the offset compensation value based on its real-time value, converting it into LED driver current output compensation instructions, pulse width modulation duty cycle compensation instructions, and color temperature and brightness adjustment parameter compensation instructions. Based on the LED driver current output compensation instructions, pulse width modulation duty cycle compensation instructions, and color temperature and brightness adjustment parameter compensation instructions, and according to the memorized reference data and the criterion compensation theory analysis... The system collects real-time LED operating data and dynamically determines the adjustment direction and magnitude through a parameter-coordinated adaptive decision-making mechanism: it adjusts the output current value of the LED driver power supply, controls the duty cycle of the pulse width modulation signal of the LED driver controller, dynamically adjusts the color temperature and brightness parameters through a digital signal processor, and, in conjunction with the group control of the LED array, allocates the output ratio of LED cool white, LED blue, LED warm white, and LED red channels and brightness levels; when the offset compensation value is greater than the offset compensation threshold, it disconnects the output of the partition relay and the electronic switch, starts the partition heat sink device, triggers the partition red alarm signal, records the risk event to the risk event database, and sends a maintenance reminder.

[0017] The second aspect of this invention provides a deep learning-driven intelligent dynamic image quality optimization system for LED light strips, including: an acquisition and preprocessing module for acquiring real-time LED operating condition data, obtaining memory benchmark data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory benchmark data, and historical fluctuation anomaly data; an intelligent spatial feature partitioning module for adaptively dividing the LED light strip into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering; a thermal drift risk classification and early warning module for analyzing the dynamic characteristics of LED light strip partitions in real time based on the time-series derivative sensitivity theory, identifying thermal drift instability and color temperature anomalies, and entering a mutation criterion process; a mutation comprehensive criterion and diversion module for performing trend mutation composite criterion on real-time LED operating condition data and memory benchmark data, executing a mutation criterion process based on the trend mutation composite criterion result, identifying and diverting chronic drift and sudden anomalies, and entering an offset compensation process; and a thermal color response offset compensation module for analyzing real-time LED operating condition data and historical fluctuation anomaly data based on criterion compensation theory, and executing an offset compensation process based on the analysis result of the criterion compensation theory.

[0018] Beneficial effects The present invention has the following beneficial effects: (1) This invention constructs a thermal drift risk classification that integrates multiple features of variance and information entropy, realizes the classification and classification of image quality anomalies and the regional security linkage, and thus achieves the effect of intelligent classification intervention and risk closed-loop management, effectively solving the problem of crude anomaly handling in the prior art.

[0019] (2) This invention proposes a dynamic mutation criterion based on nonlinear coupling and information entropy modulation, which effectively improves the sensitivity of identifying chronic drift and sudden anomalies, thereby achieving accurate identification and rapid response, and effectively solving the problem of single anomaly type identification in the prior art.

[0020] (3) This invention introduces an adaptive compensation and closed-loop adjustment mechanism, which can dynamically adjust the compensation strategy according to the results of multiple criteria, thereby achieving the effect of real-time adaptive image quality optimization and effectively solving the problem of insufficient long-term image quality consistency guarantee in the prior art.

[0021] (4) This invention, by adopting intelligent spatial feature adaptive partitioning technology, realizes precise management and independent optimization of LED light strips in different areas, thereby achieving the effect of consistent spatial distribution and uniform image quality adjustment, effectively solving the problem of insufficient overall compensation accuracy in the prior art.

[0022] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0023] Figure 1 The flowchart shows the intelligent optimization method for dynamic image quality of LED light strips based on deep learning, as described in this invention. Figure 2 This is a structural diagram of the LED light strip dynamic image quality intelligent optimization system based on deep learning driven by the present invention; Figure 3 This is a heatmap comparing the feature parameters and offset compensation values ​​of the present invention. Figure 4 This is a simplified flowchart of the thermal color response offset compensation of the present invention. Detailed Implementation

[0024] 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.

[0025] Please see Figures 1-4 This invention provides a technical solution: a method and system for intelligent dynamic image quality optimization of LED light strips based on deep learning, comprising: S1, collecting real-time LED operating condition data, acquiring memory benchmark data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory benchmark data, and historical fluctuation anomaly data; S2, adaptively dividing the LED light strip into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering; S3, analyzing the dynamic characteristics of the LED light strip partitions in real time based on the temporal derivative sensitivity theory, identifying thermal drift instability and color temperature anomalies, and entering the mutation criterion process; S4, performing trend mutation composite criterion on the real-time LED operating condition data and memory benchmark data, executing the mutation criterion process according to the trend mutation composite criterion result, identifying and diverting chronic drift and sudden anomalies, and entering the offset compensation process; S5, analyzing the real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory, and executing the offset compensation process according to the analysis result of the criterion compensation theory.

[0026] Specifically, the process of collecting real-time LED operating condition data, acquiring memory reference data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory reference data, and historical fluctuation anomaly data is as follows: real-time LED operating condition data is collected, which includes: junction temperature timing signal, main color temperature timing signal, illuminance raw data, main peak wavelength sequence, real-time power supply voltage, real-time loop current, and main color light intensity sequence. Among them, the junction temperature timing signal is collected in real time by an integrated NTC thermistor or a dedicated chip temperature sensor, and the main color temperature timing signal is acquired by a high-sensitivity color temperature sensor.

[0027] Acquire memory reference data. The memory reference data is obtained by performing standard testing procedures on the LED light strip. The memory reference data includes: memory illuminance data, memory peak data, memory power data, memory color temperature data, memory voltage, memory current, and a criterion feature sliding window. The length and step size of the criterion feature sliding window are adaptively determined according to the actual 10Hz-100Hz sampling frequency and response characteristics. The length covers 3-6 seconds of data, and the step size is 1 to 10% of the sampling frequency. The window is synchronously aligned with the physical event.

[0028] Acquire historical fluctuation and anomaly data and establish a historical fluctuation and anomaly database. The historical fluctuation and anomaly data includes: historical junction temperature fluctuation data, historical primary color temperature drift data, historical primary color deviation drift data, and historical anomaly data. The database uses a structured approach to store typical fluctuation and anomaly records for different partitions and time periods, and supports subsequent retrieval and statistical analysis by time interval, partition number, and other conditions.

[0029] Median filtering and moving average algorithms are used to suppress and smooth noise in the junction temperature time series signal, main color temperature time series signal, and raw illuminance data, eliminating short-term pulse interference and high-frequency noise during the acquisition process. The filtering window length is adaptively adjusted according to the signal characteristics, ranging from 10 to 20 points. Linear interpolation and missing value filling algorithms are used to continuously fill in short-term breakpoints and abnormal missing values ​​in the memory reference data, with no more than 5 missing points in a single instance. Batch statistics and box plot outlier removal algorithms are used to initially identify and remove outliers in historical fluctuation anomaly data, removing extreme outliers while retaining representative drift and abnormal trends. The anomaly threshold is adaptively determined based on the statistical distribution. Distribution standardization and linear normalization algorithms are used to standardize and normalize real-time LED operating condition data, memory reference data, and historical fluctuation anomaly data, eliminating dimensional differences between different physical quantities and improving the accuracy of subsequent feature extraction and criterion calculation.

[0030] This implementation plan achieves high-quality archiving and standardization of the entire process data through systematic acquisition and multi-level preprocessing of real-time LED operating condition data, memory benchmark data, and historical fluctuation and anomaly data. By leveraging sensors and platforms to acquire multi-dimensional time-series signals and combining them with intelligent algorithms such as dynamic window filtering, missing value completion, and anomaly removal, the accuracy, continuity, and representativeness of the data are effectively improved. Finally, through unified standardization and normalization processing, the plan ensures high comparability and data compatibility of various physical parameters in subsequent intelligent optimization processes such as feature extraction, criterion calculation, and adaptive compensation, laying a solid data foundation for global criterion analysis and precise image quality control.

[0031] Specifically, the process of adaptively dividing LED light strips into different independent partitions based on spatial and physical attributes through multidimensional feature analysis and clustering is as follows: LED light strips are managed by partitioning according to spatial layout and operational characteristics. Before partitioning, data modeling is performed on the overall spatial wiring structure and functional area division of the light strips. Combining the physical layout structure of the LED light strips, the grouping of driver circuits, and the thermo-optical response characteristics, the installation direction, spacing, and local heat conduction characteristics of the light strips are comprehensively considered. Spatial distance is obtained by measuring the actual physical coordinate differences. The spatial distance is extracted based on three-dimensional coordinate measurements to ensure accurate physical partitioning. Adjacency is determined based on the physical connection and grouping relationship of LED units. Adjacency is jointly determined by the driver circuit topology and series-parallel wiring information. Feature similarity is obtained by the Euclidean distance between the junction temperature timing signal, the main color temperature timing signal, and the original illuminance data. All feature data undergoes distribution standardization preprocessing to ensure that different parameters are similar. To ensure comparability between data, principal component analysis was used to reduce the dimensionality of the junction temperature time series signal, the main color temperature time series signal, and the original illuminance data. The extracted principal components were used to reflect the overall thermal and optical change trend of the region and to extract the main spatial distribution features. A spectral clustering algorithm was used in combination with spatial distance, adjacency, and feature similarity indices. The number of partitions was adaptively adjusted during the spectral clustering process to avoid subjective manual setting and to achieve multi-dimensional feature segmentation of the LED light strip. The segmentation results were corrected with the help of expert experience rules to improve the physical consistency of the partitions. The LED light strip was divided into multiple partitions, each containing several LED units with continuous structure and similar physical properties. Independent data acquisition channels and operating parameter records were established for each partition to ensure the high independence and traceability of subsequent partition risk assessment and compensation criteria.

[0032] In this implementation scheme, by modeling and intelligently partitioning the spatial structure and operational characteristics of LED light strips, a multi-dimensional parameter fusion division of spatial layout, physical connection, drive grouping, and thermo-optical response is achieved. Relying on preprocessing techniques such as 3D coordinate measurement, distribution standardization, and principal component dimensionality reduction, combined with spectral clustering and rule optimization, multiple partition units with continuous structure, similar physical properties, and consistent operational characteristics can be identified and divided with high precision. Each partition independently establishes a data acquisition channel and parameter records, enabling refined partition management of monitoring, risk assessment, and compensation criteria processes. This partitioning mechanism provides structured data support for subsequent image quality anomaly monitoring and intelligent compensation control, significantly improving the response speed, accuracy, and security of the LED light strip dynamic optimization system.

[0033] Specifically, based on the time-series derivative sensitivity theory, the real-time analysis process of the dynamic characteristics of LED strip zones is as follows: First, acquire the junction temperature time-series signal and the main color temperature time-series signal to ensure data timeliness and resolution. Second, obtain the temperature change rate of the junction temperature time-series signal using a sliding window algorithm to facilitate the capture of short-term fluctuation characteristics. Third, calculate the variance of the zone's temperature change rate using the variance within the window; this variance index reflects the dynamic activity of temperature rise changes in local areas. Fourth, obtain the mean value of the zone's temperature change rate of the junction temperature time-series signal using both the sliding window algorithm and the mean algorithm. Mean value calculation helps measure the overall level of the temperature rise trend and provides a baseline for fluctuation assessment. Fifth, analyze the main color temperature time-series signal within the sliding window. Second-order differential operations are performed to obtain the second derivative. The second-order differential is specifically used to sensitively detect abrupt inflection points in the color temperature signal. The maximum value of the absolute value of the second derivative is extracted to obtain the second-order extreme value of the color temperature shift in the partition. The extreme value reflects the maximum drastic degree of color temperature shift in the partition within a certain period of time. The color temperature shift sequence is obtained from the main color temperature time series signal through a differential algorithm. The first-order differential sequence is used to characterize the instantaneous increment of color temperature change. The standard deviation of the partition color temperature shift is obtained from the color temperature shift sequence through a sliding window standard deviation algorithm. The standard deviation is used to quantitatively describe the overall discreteness of color temperature fluctuation. The information entropy of the partition color temperature shift is obtained from the color temperature shift sequence through the Shannon entropy calculation algorithm. The information entropy reflects the complexity and uncertainty of the color temperature shift sequence.

[0034] The ratio of the variance of the temperature change rate of a zone to the sum of the absolute values ​​of the mean of the temperature change rates of the zones is used to obtain the dynamic variation coefficient of the zone temperature. This coefficient can sensitively reveal the relative relationship between local temperature rise fluctuations and steady-state levels. The ratio of the second-order extreme value of the zone color temperature shift sequence to the sum of the standard deviations of the zone color temperature shift is calculated, and multiplied by the information entropy of the zone color temperature shift to obtain the dynamic complexity of the zone color temperature, which comprehensively expresses the severity, fluctuation amplitude, and sequence complexity of the zone color temperature shift. The larger value of the dynamic variation coefficient of the zone temperature and the dynamic complexity of the zone color temperature is taken to obtain the zone thermal color deviation instability value. The specific calculation formula is as follows: ; In the formula, This indicates the thermal color deviation instability value of a zone, reflecting the degree of abnormal risk in terms of thermal drift and color temperature shift of the zone; It represents the variance of the rate of temperature change in a zone, quantifying the dynamic activity of temperature rise fluctuations in that zone; It represents the average rate of temperature change in a given area, reflecting the overall trend and steady-state level of temperature change. It represents the second-order extreme value of color temperature shift in the zone, and is sensitive to the abrupt inflection point or drastic change phenomenon in the process of color temperature change; It represents the standard deviation of color temperature offset in different zones, quantifying the dispersion and instability of color temperature changes; It represents the entropy of color temperature offset information in different zones, quantifying the dispersion and instability of color temperature changes; This represents the temperature compensation constant, which is obtained by using a robust constraint algorithm based on the minimum value of the denominator for the mean rate of temperature change in the partition. Its value ranges from 0.005 to 0.01. This represents the color temperature compensation constant, which is obtained by using a robust constraint algorithm based on the minimum value of the denominator for the standard deviation of color temperature offset in different zones. The value ranges from 0.005 to 0.01.

[0035] This implementation scheme can accurately identify the risks of thermal drift and color temperature shift instability in each zone of an LED light strip under complex operating conditions, enabling early warning and quantitative classification of regional anomalies. The processing not only comprehensively considers core factors such as temperature rise activity, color temperature abrupt changes, fluctuation amplitude, and complexity of changes, but also utilizes an adaptive compensation constant to ensure the robustness of the criteria in low-signal-fluidity scenarios. The final obtained zone thermal-color shift instability values ​​provide an efficient and quantifiable basis for subsequent warning zone selection, graded response, and dynamic compensation, effectively improving the LED image quality intelligent optimization system's ability to quickly perceive and finely control thermal-color dynamic anomalies.

[0036] Specifically, the process of identifying thermal drift instability and color temperature anomalies and entering the mutation criterion process is as follows: real-time comparison of the zoned thermal color deviation instability value and the zoned thermal color deviation instability threshold to ensure real-time performance and stability. The zoned thermal color deviation instability threshold includes a primary instability threshold and a secondary instability threshold, supporting dynamic adjustment at the zone level. When the thermal color deviation instability value of a zone is less than the secondary instability threshold, it is determined to be a safe zone. The zone operates under normal current conditions and requires no special intervention, ensuring the normal operation of the zone.

[0037] When the thermal color deviation instability value of a zone is greater than or equal to the second-level instability threshold but less than the first-level instability threshold, it is determined to be a warning zone. The LED drive current is dynamically reduced by the digital signal processor to suppress the continued accumulation of heat. The sampling frequency of real-time LED operating condition data is increased to improve the ability to detect anomalies in a finer granular manner. All zone sampling data and response measures are recorded, and the zone enters the mutation criterion process.

[0038] When the thermal color deviation instability value of a zone is greater than or equal to the first-level instability threshold, it is identified as a risk zone. The zone relay output and electronic switch are disconnected, the working circuit of the risk zone is forcibly cut off to ensure safety, the zone heat sink device is activated, the hardware active cooling mechanism is linked, and a red alarm signal for the zone is triggered. The zone alarm signal is uploaded to the centralized monitoring terminal and the operation and maintenance backend in real time. At the same time, the risk event is created and recorded in the risk event database, and the trigger time, zone number and main abnormal parameters are archived in detail to facilitate subsequent traceability analysis and optimization. Maintenance reminders are sent to realize a closed loop of remote intelligent early warning and hierarchical response.

[0039] This implementation scheme achieves real-time quantification and intelligent classification of the status of each physical zone of the LED light strip in terms of thermal drift and color temperature anomalies through a zone-level discrimination and dynamic hierarchical response mechanism for thermal color deviation instability values. The safe zone maintains stable operation, while the warning zone can promptly reduce the drive current, increase the sampling frequency of operating conditions, and comprehensively record the details of zone anomalies without interfering with the main process, effectively preventing further heat accumulation and anomaly propagation. The risk zone forcibly cuts off the power supply to the abnormal zone through hardware linkage, coordinating with heat sink devices and alarm mechanisms to ensure the safety of the LED strip itself and downstream equipment. The entire process of risk events in each zone is archived and synchronized in real time with the operation and maintenance backend, providing detailed data support for subsequent accurate tracing and full lifecycle optimization. Through multi-level discrimination, precise intervention, and closed-loop archiving, this module significantly improves the image quality stability and operational safety of LED light strips in complex application environments, laying the core foundation for intelligent optimization and adaptive maintenance.

[0040] Specifically, the process of applying a trend change composite criterion to real-time LED operating condition data and memory reference data is as follows: Acquiring raw illuminance data, memory illuminance data, peak wavelength sequence, memory peak data, main color temperature timing signal, memory color temperature data, real-time power supply voltage, memory voltage, real-time loop current, and memory current constitutes the multi-dimensional input reference for the trend change criterion; calculating the difference between the acquired raw illuminance data and the memory illuminance data using a sliding window algorithm yields the zonal brightness offset; the sliding window mechanism effectively smooths out instantaneous fluctuation interference; calculating the difference between the peak wavelength sequence and the memory peak data using a sliding window algorithm yields the peak wavelength offset. The system employs several methods: 1) Shift values: Continuous sliding comparison captures wavelength gradient trends; 2) Differential calculation of the main color temperature time-series signal and memory color temperature data yields color temperature offset values, enabling rapid response to dynamic color temperature changes; 3) Differential calculation of the real-time supply voltage and memory voltage of each zone yields voltage offset values, allowing real-time voltage comparison to monitor power supply stability; 4) Differential calculation of the real-time loop current and memory current of each zone yields current offset values, enabling current difference analysis to assess the operating status of the drive circuit; 5) Multiplication of the real-time supply voltage and real-time loop current yields real-time power, and power product calculation accurately reflects instantaneous energy consumption levels. The difference between real-time power and memory power data is then used to obtain the power offset value.

[0041] The product of luminance offset and peak wavelength offset is calculated to characterize the coupling fluctuation relationship between luminance and wavelength parameters. The variance of this product within the criterion feature sliding window is statistically analyzed to quantify the severity of the coupling fluctuation. The absolute value of the mean of the criterion feature sliding window for power offset is incremented by one to avoid zero denominators and smooth the calculation, resulting in the power mean term. The variance result is divided by the power mean term, and the square root is taken. The square root operation restores the dimensions and weakens the influence of extreme values, resulting in the luminance-wavelength joint fluctuation term. The product of color temperature offset and voltage offset sine is calculated, and a sine function is introduced to map the voltage offset to a periodic sensitive interval. The derivative of the mean of the criterion feature sliding window with respect to time is calculated. The time derivative, with the sampling period as the basic granularity, is used to capture the dynamic change trend of the parameters. This derivative is multiplied by the trend change sensitivity factor. The weights for amplifying and reducing trend changes are determined by taking their absolute values ​​to ensure the response term is always positive, resulting in the trend change response term. The difference between the maximum and minimum values ​​of the product of the main peak wavelength offset and the current offset within the criterion feature sliding window is calculated. The standard deviation of the criterion feature sliding window for the brightness offset is incremented by one to prevent division by zero and increase computational stability. The range result is divided by the standard deviation term and multiplied by the mutation response adjustment factor to control the response intensity to abnormal fluctuations. This factor serves as the base of the exponent, ensuring the monotonic growth characteristic of the exponential function. When the calculated base value is less than 1, it is corrected to 1. The exponential function value is then used to nonlinearly amplify significant anomalies, resulting in the mutation range amplification term. The brightness wavelength joint fluctuation term, the trend change response term, and the mutation range amplification term are added together to obtain the mutation criterion value. The specific calculation formula is as follows: ; In the formula, This represents the mutation criterion value, which comprehensively reflects the dynamic characteristics of the joint offset of partitioned physical quantities and abnormal mutations; This indicates the brightness offset value, reflecting the degree of difference between the current brightness and the standard operating conditions; It represents the main peak wavelength shift value, depicting the trend of LED light emission main color shift; This indicates the power offset value, which visually reflects changes in LED load and power fluctuations. This indicates the color temperature offset value, which measures the degree of color temperature fluctuation and color consistency. It represents the voltage offset value, revealing the dynamic fluctuations and transient anomalies of the power supply system; It indicates the current offset value, and sensitively detects abnormal output and device failure risk in the drive circuit; The trend change sensitivity factor is obtained by using the empirical mode decomposition (EMD) algorithm on historical voltage offset values ​​and historical color temperature offset values. The value ranges from 0.1 to 2.0. The mutation response adjustment factor is obtained by using a self-learning algorithm based on the historical peak wavelength offset and historical current offset values. The value ranges from 0.05 to 0.5.

[0042] In this implementation scheme, intelligent diagnosis of the LED light strip's operating status is achieved by calculating multi-parameter offsets to obtain real-time deviation data across multiple dimensions, including brightness, wavelength, color temperature, voltage, current, and power. Based on this, three core feature terms are innovatively constructed: a brightness-wavelength joint fluctuation term, processed through variance analysis and power normalization, effectively captures coordinated anomalies in optical parameters; a trend change response term, utilizing derivative calculations and sensitivity adjustment, accurately identifies gradual trends in parameters; and a sudden change range amplification term, through range statistics and exponential amplification mechanisms, enhances the response capability to sudden anomalies. Finally, through the linear fusion of these three features, a sudden change criterion value that comprehensively characterizes the state is generated, providing a reliable quantitative basis for subsequent identification of chronic drift and sudden anomalies, achieving a technological leap from simple threshold judgment to multi-feature fusion intelligent diagnosis.

[0043] Specifically, the process of executing the mutation criterion procedure based on the trend mutation composite criterion results, identifying and triaging chronic drift and sudden anomalies, and then proceeding to the offset compensation procedure is as follows: The mutation criterion value and the mutation criterion threshold are compared in real time, and the state is determined by the threshold boundary. The mutation criterion process is then executed.

[0044] When the mutation criterion value is less than the mutation criterion threshold, it indicates that the chip is in a slow change stage and is identified as a chronic drift region, where the chip junction temperature rises. The drift data and the entire process of partition drift are recorded, a chronic drift log is generated, the downstream compensation processing module is activated, and the offset compensation process begins.

[0045] When the mutation criterion value is greater than or equal to the mutation criterion threshold, it indicates that an abnormal mutation exceeding the normal range has been detected, and it is determined to be a sudden abnormal zone. Hardware-level security isolation is initiated, the LED drive circuit is immediately disconnected, the partition security lock is initiated, the partition red alarm signal is triggered, the sudden abnormal event is recorded to the risk event database, and a maintenance reminder is sent.

[0046] In this implementation plan, this step, based on the quantitative analysis of the mutation criterion value, achieves accurate diagnosis and intelligent current distribution of the LED light strip's operating status. By comparing the real-time calculated mutation criterion value with the mutation criterion threshold, it is possible to effectively distinguish between two different fault modes: chronic drift and sudden anomaly. For cases determined to be chronic drift, while recording complete drift data, a downstream compensation mechanism is activated to proactively compensate for performance degradation. For sudden anomalies, multiple protection measures, including drive circuit disconnection, safety locking, and alarm notification, are initiated to effectively prevent the fault from escalating. This hierarchical processing mechanism ensures both continuous and stable operation and rapid and safe response in truly dangerous situations, forming a complete closed-loop control system for monitoring, diagnosis, and processing.

[0047] Specifically, the process of analyzing real-time LED operating condition data and historical fluctuation anomaly data based on criterion compensation theory is as follows: Acquire junction temperature timing signals, primary color temperature timing signals, primary peak wavelength sequences, primary color intensity sequences, historical junction temperature fluctuation data, historical primary color temperature drift data, historical primary color deviation drift data, and historical anomaly data, ensuring data integrity and timing synchronization; verify the primary peak wavelength sequences and primary color intensity sequences through CIE... The 1976 uniform chromaticity space transformation algorithm obtains the primary color deviation time-series signal, realizing instantaneous feature mapping and color deviation quantization of multicolor light signals. The cross-correlation information entropy is obtained from the junction temperature time-series signal and the primary color temperature time-series signal using a sliding window cross-correlation analysis algorithm and a Shannon entropy calculation algorithm. The sliding window length is adaptive to the sampling frequency, and the thermal-color coupling strength and complexity are dynamically evaluated in conjunction with the real-time signal stream. The junction temperature and primary color temperature coupling characteristic quantities are obtained from the junction temperature and primary color temperature time-series signals using a sliding window cross-correlation analysis algorithm, quantitatively reflecting the coupling driving effect of temperature changes on primary color temperature fluctuations and capturing the trend of thermal-color co-evolution. The junction temperature and primary color deviation coupling characteristic quantities are also obtained from the junction temperature and primary color deviation time-series signals using a sliding window cross-correlation analysis algorithm, used to measure the strong correlation between abnormal junction temperature fluctuations and abrupt changes in primary color deviation, assisting in the identification of instability precursors.

[0048] The product of the junction temperature response enhancement factor raised to the power of the junction temperature change rate and the primary color temperature response enhancement factor raised to the power of the primary color temperature change rate yields the thermochromic response enhancement term, reflecting the nonlinear amplification characteristics of the local response. The primary color bias response suppression factor raised to the power of the primary color bias change rate plus one yields the primary color bias suppression term, which dynamically adjusts the sensitivity to color bias fluctuations and effectively suppresses the influence of irrelevant disturbances on the criterion. Dividing the thermochromic response enhancement term by the primary color bias suppression term yields the coupling ratio term, which comprehensively reflects the dynamic criterion of thermochromic coupling and primary color bias stability. The mutual information entropy of the junction temperature and primary color temperature coupling characteristics is calculated to obtain the coupling information entropy term, characterizing the complexity and non-stationary risk of the coupling structure. The square of the junction temperature and primary color bias coupling characteristics is multiplied by the abnormal activation adjustment factor to calculate the exponential function, yielding the abnormal activation amplification term, which amplifies the impact of extreme abnormal events and strengthens the response capability to sudden instability. Multiplying the coupling ratio term, coupling information entropy term, and abnormal activation amplification term yields the offset compensation value, calculated using the following formula: ; In the formula, This represents the offset compensation value, which characterizes the magnitude and direction of the dynamic compensation required in cases of thermal color drift and color shift anomalies. This represents the junction temperature timing signal, reflecting the real-time temperature rise and dynamic trend of the LED chip junction region; It represents the primary color temperature timing signal, dynamically tracks the real-time changes in the primary color temperature of the LED output spectrum, and is used to determine color temperature drift and instability behavior; It represents the primary color shift timing signal, reflecting the real-time color shift dynamics under the superposition of different color channels, and providing a quantitative basis for color consistency compensation; Represents the information entropy of mutual correlation, effectively identifying thermal color synergy anomalies and nonlinear risks; This represents the junction temperature and primary color temperature coupling characteristic quantities, and the trend and potential driving force of thermally induced color temperature instability in sensitive detection systems; It represents the junction temperature primary color bias coupling characteristic quantity, used to identify the color bias linkage effect induced by thermal drift and the precursors of instability; The junction temperature response enhancement factor is defined as follows: the junction temperature change rate is extracted from the junction temperature time series signal using a sliding window differential algorithm; anomaly sensitivity backtracking analysis is performed on the junction temperature change rate and historical junction temperature fluctuation data; and the junction temperature response enhancement factor is obtained by combining anomaly distribution characteristics optimization algorithm. The value range is between 1.0 and 2.5. The primary color temperature response enhancement factor is defined as follows: the primary color temperature change rate is extracted from the primary color temperature time series signal using a sliding window differential algorithm; anomaly sensitivity backtracking analysis is performed on the primary color temperature change rate and historical primary color temperature drift data; and the primary color temperature response enhancement factor is obtained by combining the optimal fitting and tuning algorithm, with a value range between 1.0 and 2.5. The main color deviation response suppression factor is represented by the sliding window differential algorithm to extract the main color deviation change rate from the main color deviation time series signal. Anomaly sensitivity backtracking analysis is performed on the main color deviation change rate and historical main color deviation drift data. The main color deviation response suppression factor is obtained by combining the historical color deviation fluctuation analysis algorithm. The value range is between 0.5 and 2.0. The abnormal activation adjustment factor is obtained by using clustering and statistical learning algorithms and multivariate adjustment sensitivity evaluation algorithms on junction temperature time series signals, primary color temperature time series signals, primary color deviation time series signals and historical abnormal data. The abnormal activation adjustment factor ranges from 0.01 to 1.0.

[0049] In this embodiment, Table 1 is a comparison table of characteristic parameters and offset compensation values. It records in detail the junction temperature change rate, main color temperature change rate, main color deviation change rate, junction temperature and main color temperature coupling characteristic quantity, junction temperature and main color deviation coupling characteristic quantity, mutual related information entropy, and finally calculated offset compensation value of different partitions under typical working conditions. It is used to quantify the comprehensive performance of multi-physical quantity coupling and adaptive compensation under the working state of LEDs in each partition. Among them, the junction temperature change rate of serial number 1 is 0.20, the primary color temperature change rate is 0.35, the primary color deviation change rate is 0.12, the junction temperature and primary color temperature coupling characteristic is 0.65, the junction temperature and primary color deviation coupling characteristic is 0.25, the cross-correlation entropy is 0.72, and the compensation value is 0.35; the compensation value of serial number 3 is the highest at 0.81, and the compensation value of serial number 5 is the lowest at 0.39, reflecting the difference in the response ability of different partitions to temperature and light color changes. The compensation sensitivity parameter q and the primary color temperature weight w, which are not mentioned in the table, are fixed at 1.5 and 0.8, respectively, to eliminate the influence of fluctuations in other variables and ensure the comparability of the main characteristic parameters with the compensation values.

[0050] Table 1 Comparison of Feature Parameters and Offset Compensation Values

[0051] like Figure 3 The heatmap shown compares the characteristic parameters with the offset compensation values, intuitively reflecting the parameter distribution of different zones in terms of junction temperature change, primary color temperature change, primary color shift change, and related characteristic quantities, and their impact on the final offset compensation value. As can be seen from the data in Table 1, the characteristic parameters of number 3 are generally higher, and the corresponding offset compensation value is also the highest, indicating its strongest adaptive adjustment capability under complex operating conditions. The main characteristic parameters of number 5 are generally lower, and the corresponding compensation value is also at the lowest level, indicating its relatively stable operation. Overall, the heatmap demonstrates the dynamic relationship between the characteristic parameters of each zone and the compensation criteria, providing data support for zone optimization, anomaly warning, and refined compensation strategies.

[0052] In this implementation plan, this step systematically acquires, normalizes, and dynamically calculates multiple feature quantities based on key physical parameters during LED strip operation, forming a multi-dimensional feature parameter system reflecting the complex state changes of different zones. Through the organic combination of feature parameters and compensation values, the response differences of each zone to changes in temperature, color temperature, and color shift under different operating conditions can be intuitively quantified, revealing their dynamic impact on overall compensation requirements. By comprehensively employing data fusion, feature mapping, and coupled criterion methods, the system not only enhances the ability to identify thermal color anomalies, drift trends, and local anomalous changes, but also provides a highly usable feature data foundation for the implementation of subsequent LED drive compensation, risk protection, and intelligent control strategies. This step improves the LED strip system's adaptive perception capability and data support level for complex operating states, laying a solid foundation for intelligent image quality optimization and safe operation.

[0053] Specifically, the process of executing the offset compensation procedure based on the analysis results of the criterion compensation theory is as follows: The offset compensation value and the offset compensation threshold are compared in real time, and the offset compensation procedure is executed. When the offset compensation value is less than or equal to the offset compensation threshold, a sigmoid function is used for mapping based on the real-time value of the offset compensation value. The output range of the sigmoid function for compensation mapping is strictly limited to the upper and lower limits of the physical safety of each instruction. The adjustment ranges of current, duty cycle, color temperature, and brightness are limited to ±20%, ±15%, ±10%, and ±10% of the memory value, respectively. The mapping slope is preferably between 3 and 8. The actual value is determined by the current response time constant and heat dissipation performance results corresponding to each LED channel. The adjustment rate is constrained by the physical inertia of the LED device and the system safety margin. The parameters can be adaptively optimized according to the equipment type and historical monitoring data. The offset compensation value is converted into LED driver current output compensation instructions, pulse width modulation duty cycle compensation instructions, and color temperature and brightness adjustment parameter compensation instructions. Based on the LED driver's current output compensation command, pulse width modulation duty cycle compensation command, and color temperature and brightness adjustment parameter compensation command, and according to the real-time LED operating condition data involved in the analysis of the memory reference data and the criterion compensation theory, the adjustment direction and amplitude are dynamically determined through the parameter collaborative adaptive decision-making mechanism. By comparing the memory reference data with the real-time LED operating condition data, and combining the parameter change rate, a dynamic rate limiting gating mechanism is adopted to impose hard constraints on the compensation step and maximum adjustment amplitude of each control channel, and to determine whether there is overcompensation, rapid reversal, and signal oscillation trend. Once the constraint threshold is triggered, a soft limiting algorithm is activated to suppress the adjustment rate and perform compensation buffering: the output current value of the LED driver power supply is adjusted, and a real-time PI adaptive PID closed-loop algorithm is used to smooth the current deviation. An output buffer is introduced into the control process to prevent current sudden changes and short-term overshoot. The duty cycle of the pulse width modulation signal of the LED driver controller is controlled, and the color temperature and brightness parameters are dynamically adjusted through the digital signal processor. Combined with the group control of the LED array, the output ratio of LED cool white, LED blue, LED warm white, and LED red channels and brightness levels is allocated.

[0054] When the offset compensation value exceeds the offset compensation threshold, the partition relay output and electronic switch are disconnected, the partition heat sink device is activated, and the forced power reduction and active ventilation of the high-temperature area are linked to ensure timely release of local heat, triggering the partition red alarm signal. At the same time, the risk event is recorded to the risk event database, maintenance reminders are sent, and operation and maintenance notifications are pushed to the back-end management terminal to ensure timely response and closed-loop tracking of on-site maintenance.

[0055] like Figure 4 The diagram shows a simplified flowchart of thermal color response offset compensation, illustrating the core logic of the thermal color response offset compensation module: First, real-time LED operating condition data, memory reference data, and historical fluctuation anomaly data are input. After the compensation value is calculated, based on the comparison between the offset compensation value and the threshold, two paths are entered: parameter adaptive compensation and high-risk emergency handling. All branches are monitored and fed back in a closed loop, realizing dynamic adaptive adjustment of the LED light strip's operating status and risk linkage protection.

[0056] In this implementation plan, this step relies on strictly defined physical upper and lower limits and dynamic adjustment rates, combined with real-time LED operating data and memorized reference data, to implement smooth and gradual compensation adjustments to core output parameters such as drive current, pulse width modulation duty cycle, color temperature, and brightness through a parameter collaborative adaptive mechanism. By integrating the results of criterion-based compensation theory analysis, the adjustment direction and amplitude of each channel are dynamically determined, and overcompensation, signal oscillation, and sudden change risks are detected and controlled in real time, ensuring that compensation commands always operate efficiently within safety margins and physical limitations. If the compensation threshold is exceeded, a multi-level linkage safety strategy can be quickly switched to forced power-off, active heat dissipation, and red risk warning, forming a closed-loop control system from fine-tuning compensation to extreme protection, significantly improving the adaptive image quality adjustment capability and operational safety assurance level of the LED light strip in complex environments.

[0057] like Figure 2 As shown, the second aspect of this invention provides a deep learning-driven intelligent dynamic image quality optimization system for LED light strips, including an acquisition and preprocessing module for acquiring real-time LED operating condition data, obtaining memory benchmark data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory benchmark data, and historical fluctuation anomaly data; an intelligent spatial feature partitioning module for adaptively dividing the LED light strip into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering; a thermal drift risk classification and early warning module for analyzing the dynamic characteristics of LED light strip partitions in real time based on the time-series derivative sensitivity theory, identifying thermal drift instability and color temperature anomalies, and entering the mutation criterion process; a mutation comprehensive criterion and diversion module for performing trend mutation composite criterion on real-time LED operating condition data and memory benchmark data, executing the mutation criterion process according to the trend mutation composite criterion result, identifying and diverting chronic drift and sudden anomalies, and entering the offset compensation process; and a thermal color response offset compensation module for analyzing real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory, and executing the offset compensation process according to the analysis result of the criterion compensation theory.

[0058] This implementation scheme achieves intelligent optimization management of the entire process of LED light strip image quality status through multi-level module collaboration. Each functional module supports the others. High-precision data acquisition and preprocessing at the front end lays the data foundation for subsequent intelligent zoning, risk identification, and anomaly compensation. The intelligent spatial feature zoning module ensures independent and targeted risk assessment and control strategies for each physical area. Built-in thermal drift risk classification and early warning can keenly capture and respond to zoned thermal color deviation instability and color temperature anomalies; the sudden change comprehensive criterion and diversion process achieve accurate identification and response diversion for chronic and sudden anomalies. Finally, the thermal color response offset compensation module integrates multi-dimensional criteria with historical experience, dynamically adjusts driving parameters, and forms a closed-loop adaptive image quality optimization system. This improves the adaptive image quality management and anomaly prevention capabilities of the LED light strip system under complex operating conditions, effectively ensuring image quality consistency and operational safety during long-term operation.

[0059] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0060] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for intelligent dynamic image quality optimization of LED light strips based on deep learning, characterized in that, Includes the following steps: S1, Collect real-time LED operating condition data, obtain memory reference data and historical fluctuation anomaly data, and preprocess the real-time LED operating condition data, memory reference data and historical fluctuation anomaly data; S2, through multidimensional feature analysis and clustering, adaptively divides LED light strips into different independent partitions according to spatial and physical attributes; S3, based on the time derivative sensitivity theory, analyzes the dynamic characteristics of LED light strip zones in real time, identifies thermal drift instability and color temperature abnormalities, and enters the mutation criterion process; S4 performs a trend change composite criterion on real-time LED operating condition data and memory reference data, executes the change criterion process based on the trend change composite criterion result, identifies and diverts chronic drift and sudden anomalies, and enters the offset compensation process. S5 analyzes real-time LED operating data and historical fluctuation anomaly data based on the criterion compensation theory, and executes the offset compensation process according to the analysis results of the criterion compensation theory.

2. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of collecting real-time LED operating condition data, obtaining memory reference data and historical fluctuation anomaly data, and preprocessing the real-time LED operating condition data, memory reference data, and historical fluctuation anomaly data is as follows: Collect real-time LED operating condition data, including: junction temperature timing signal, main color temperature timing signal, raw illuminance data, main peak wavelength sequence, real-time power supply voltage, real-time loop current, and main color light intensity sequence; Acquire memory reference data, which includes: memory illuminance data, memory main peak data, memory power data, memory color temperature data, memory voltage, memory current, and a sliding window of criteria features; Acquire historical fluctuation anomaly data and establish a historical fluctuation anomaly database. The historical fluctuation anomaly data includes: historical junction temperature fluctuation data, historical primary color temperature drift data, historical primary color deviation drift data, and historical anomaly data. Median filtering and moving average algorithms are used to suppress and smooth noise in the junction temperature time series signal, main color temperature time series signal, and raw illuminance data, eliminating short-term pulse interference and high-frequency noise during the acquisition process. Linear interpolation and missing value filling algorithms are used to continuously fill in short-term breakpoints and abnormal missing values ​​in the memory reference data. Batch statistics and box plot outlier removal algorithms are used to initially identify and remove outliers in historical fluctuation data, removing extreme outliers and retaining representative drift and abnormal trends. Distribution standardization and linear normalization algorithms are used to standardize and normalize real-time LED operating condition data, memory reference data, and historical fluctuation data.

3. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of adaptively dividing LED light strips into different independent partitions based on spatial and physical attributes through multidimensional feature analysis and clustering is as follows: The LED light strips are managed by partitioning according to their spatial layout and operating characteristics. Combining the physical arrangement structure, driving circuit grouping, and thermo-optical response characteristics of the LED light strips, spatial distance is obtained by measuring the difference in actual physical coordinates. Adjacency is determined based on the physical connection and grouping relationship of LED units. Feature similarity is obtained by the Euclidean distance between the junction temperature time-series signal, the main color temperature time-series signal, and the original illuminance data. Principal component analysis is used to perform dimensionality reduction processing on the junction temperature time-series signal, the main color temperature time-series signal, and the original illuminance data to extract the main spatial distribution features. Then, a spectral clustering algorithm is used to combine spatial distance, adjacency, and feature similarity indexes to achieve multi-dimensional feature segmentation of the LED light strips. The LED light strips are divided into multiple partitions, each containing several LED units with continuous structure and similar physical properties. Each partition establishes an independent data acquisition channel and operating parameter record.

4. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of analyzing the dynamic characteristics of LED strip zones in real time based on the time-series derivative sensitivity theory is as follows: Acquire the junction temperature time-series signal and the main color temperature time-series signal; obtain the temperature change rate of the junction temperature time-series signal using a sliding window algorithm, and calculate the variance of the temperature change rate of each zone using the variance within the window; obtain the mean of the temperature change rate of each zone using a sliding window algorithm and a mean algorithm; obtain the second derivative by performing a second-order difference operation on the main color temperature time-series signal within the sliding window, and extract the maximum value of the absolute value of the second derivative to obtain the second-order extreme value of the zone color temperature offset; obtain the color temperature offset sequence from the main color temperature time-series signal using a difference algorithm, and obtain the standard deviation of the zone color temperature offset using a sliding window standard deviation algorithm; obtain the zone color temperature offset information entropy using the Shannon entropy calculation algorithm for the color temperature offset sequence. The ratio of the variance of the temperature change rate of a zone to the sum of the absolute values ​​of the mean of the temperature change rates of the zones is used to obtain the dynamic variation coefficient of the zone temperature. The ratio of the second extreme value of the color temperature shift sequence of a zone to the sum of the standard deviation of the color temperature shift of the zone is calculated and multiplied by the entropy of the color temperature shift of the zone to obtain the dynamic complexity of the zone color temperature. The larger value of the dynamic variation coefficient of the zone temperature and the dynamic complexity of the zone color temperature is taken to obtain the instability value of the zone thermal color deviation.

5. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process for identifying thermal drift instability and color temperature anomalies, and then proceeding to the mutation criterion process, is as follows: Real-time comparison of the partitioned thermal color deviation instability value and the partitioned thermal color deviation instability threshold. The partitioned thermal color deviation instability threshold includes a primary instability threshold and a secondary instability threshold. When the thermal color deviation instability value of the partition is less than the second-level instability threshold, it is determined to be a safe zone, and the partition operates in a normal current state without special intervention. When the thermal color deviation instability value of a zone is greater than or equal to the second-level instability threshold and less than the first-level instability threshold, it is determined to be a warning zone. The LED driving current is reduced, the sampling frequency of real-time LED operating condition data is increased, and the zone enters the sudden change criterion process. When the thermal color deviation instability value of a zone is greater than or equal to the first-level instability threshold, it is determined to be a risk zone. The zone relay output and electronic switch are disconnected, the zone heat sink device is activated, the zone red alarm signal is triggered, and a risk event is created and recorded in the risk event database, and a maintenance reminder is sent.

6. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of applying a composite criterion for trend change in real-time LED operating condition data and memory baseline data is as follows: Acquire raw illuminance data, memorized illuminance data, peak wavelength sequence, memorized peak data, primary color temperature time-series signal, memorized color temperature data, real-time power supply voltage, memorized voltage, real-time loop current, and memorized current; calculate the difference between the acquired raw illuminance data and the memorized illuminance data using a sliding window algorithm to obtain the partition brightness offset; calculate the difference between the peak wavelength sequence and the memorized peak data using a sliding window algorithm to obtain the peak wavelength offset value. The color temperature offset value is obtained by calculating the difference between the primary color temperature timing signal and the memory color temperature data. The voltage offset value is obtained by calculating the difference between the real-time power supply voltage and the memory voltage of the partition; The current offset value is obtained by calculating the difference between the real-time loop current and the memory current of the partition. The real-time power is obtained by multiplying the real-time supply voltage and the real-time loop current. The power offset value is obtained by calculating the difference between the real-time power and the stored power data. Calculate the product of the luminance offset and the main peak wavelength offset, calculate the variance of this product within the criterion feature sliding window, calculate the absolute value of the mean of the criterion feature sliding window for the power offset and add one to obtain the power mean term, divide the variance result by the power mean term, and take the arithmetic square root to obtain the luminance-wavelength joint fluctuation term; calculate the product of the color temperature offset and the sine of the voltage offset, find the derivative of the mean of the criterion feature sliding window with respect to time, multiply it by the trend change sensitivity factor, and take the absolute value to obtain the trend change response term; Calculate the difference between the maximum and minimum values ​​of the product of the main peak wavelength offset and the current offset within the criterion feature sliding window. Calculate the standard deviation of the criterion feature sliding window for the brightness offset, add one, divide the range result by the standard deviation term, multiply by the abrupt change response adjustment factor, use this as the base of the exponent, and take the exponential function value to obtain the abrupt change range amplification term. Add the brightness wavelength joint fluctuation term, the trend change response term, and the abrupt change range amplification term to obtain the abrupt change criterion value.

7. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of executing the mutation criterion process based on the trend mutation composite criterion result, identifying and triaging chronic drift and sudden anomalies, and then entering the offset compensation process is as follows: Real-time comparison of mutation criterion values ​​and mutation criterion thresholds, and execution of the mutation criterion process: When the mutation criterion value is less than the mutation criterion threshold, it is determined to be a chronic drift region, and there is a chip junction temperature rise phenomenon; record the drift data and the entire process of partition drift, generate a chronic drift log, and enter the offset compensation process; When the mutation criterion value is greater than or equal to the mutation criterion threshold, it is determined to be a sudden anomaly zone; Immediately disconnect the LED drive circuit, activate the partition safety lock, trigger the partition red alarm signal, record the sudden abnormal event to the risk event database, and send a maintenance reminder.

8. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of analyzing real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory is as follows: Acquire junction temperature time-series signal, primary color temperature time-series signal, primary peak wavelength sequence, primary color intensity sequence, historical junction temperature fluctuation data, historical primary color temperature drift data, historical primary color deviation drift data, and historical anomaly data; obtain the primary color deviation time-series signal from the primary peak wavelength sequence and primary color intensity sequence using the CIE 1976 uniform color space transformation algorithm; obtain the cross-correlation entropy between the junction temperature time-series signal and the primary color temperature time-series signal using the sliding window cross-correlation analysis algorithm and the Shannon entropy calculation algorithm; obtain the junction temperature and primary color temperature coupling characteristic quantity from the junction temperature time-series signal and the primary color temperature time-series signal using the sliding window cross-correlation analysis algorithm; obtain the junction temperature and primary color deviation coupling characteristic quantity from the junction temperature time-series signal and the primary color deviation time-series signal using the sliding window cross-correlation analysis algorithm. The thermochromic response enhancement term is obtained by multiplying the junction temperature response enhancement factor raised to the power of the junction temperature change rate by the primary color temperature response enhancement factor raised to the power of the primary color temperature change rate. The main color shift response suppression factor is calculated by raising it to the power of the main color shift change rate and then adding one to obtain the main color shift suppression term. Divide the thermal color response enhancement term by the primary color bias suppression term to obtain the coupling ratio term; The cross-correlation entropy of the junction temperature and main color temperature coupled features is calculated to obtain the coupling information entropy term; The exponential function is calculated by multiplying the square of the junction temperature primary color bias coupling characteristic by the abnormal activation adjustment factor to obtain the abnormal activation amplification term; the offset compensation value is obtained by multiplying the coupling ratio term, the coupling information entropy term, and the abnormal activation amplification term.

9. The method for intelligent dynamic image quality optimization of LED light strips based on deep learning as described in claim 1, characterized in that: The specific process of performing the offset compensation procedure based on the analysis results of the criterion compensation theory is as follows: Real-time comparison of offset compensation values ​​and offset compensation thresholds, and execution of offset compensation procedures: When the offset compensation value is less than or equal to the offset compensation threshold, the sigmoid function is used to map the offset compensation value based on the real-time value, converting it into LED driver current output compensation command, pulse width modulation duty cycle compensation command, and color temperature and brightness adjustment parameter compensation command. Based on the LED driver current output compensation command, pulse width modulation duty cycle compensation command, color temperature and brightness adjustment parameter compensation command, and real-time LED operating condition data involved in the analysis of memory reference data and criterion compensation theory, the adjustment direction and amplitude are dynamically determined through parameter collaborative adaptive decision-making mechanism: adjusting the output current value of LED driver power supply, controlling the pulse width modulation signal duty cycle of LED driver controller, dynamically adjusting color temperature and brightness parameters through digital signal processor, and combining the group control of LED array to allocate the output ratio of LED cool white, LED blue light, LED warm white and LED red light channels and brightness levels; When the offset compensation value exceeds the offset compensation threshold, the partition relay output and electronic switch are disconnected, the partition heat sink device is activated, the partition red alarm signal is triggered, the risk event is recorded to the risk event database, and a maintenance reminder is sent.

10. A deep learning-driven intelligent dynamic image quality optimization system for LED light strips, employing the deep learning-driven intelligent dynamic image quality optimization method for LED light strips as described in any one of claims 1-9, comprising: The data acquisition and preprocessing module is used to acquire real-time LED operating condition data, obtain memory reference data and historical fluctuation anomaly data, and preprocess the real-time LED operating condition data, memory reference data and historical fluctuation anomaly data. The intelligent spatial feature partitioning module is used to adaptively divide LED light strips into different independent partitions according to spatial and physical attributes through multi-dimensional feature analysis and clustering. The thermal drift risk classification and early warning module is used to analyze the dynamic characteristics of LED light strip zones in real time based on the time derivative sensitivity theory, identify thermal drift instability and color temperature abnormalities, and enter the mutation criterion process. The mutation comprehensive criterion and diversion module is used to perform trend mutation composite criterion on real-time LED operating condition data and memory reference data. Based on the trend mutation composite criterion result, the mutation criterion process is executed to identify and divert chronic drift and sudden anomalies, and then enter the offset compensation process. The thermal color response offset compensation module is used to analyze real-time LED operating condition data and historical fluctuation anomaly data based on the criterion compensation theory, and execute the offset compensation process according to the analysis results of the criterion compensation theory.

Images