Display screen brightness control method for intelligent storage cabinet

Abstract

The invention discloses a display screen brightness control method for an intelligent storage cabinet. The method comprises the following steps: acquiring environment illumination data and time data acquired by a plurality of illumination sensors to generate an environment feature vector; constructing a brightness mapping model by using a neural evolutionary algorithm to obtain an initial brightness value; brightness optimization is carried out by adopting a constraint optimization method based on an Ising machine; and dynamically optimizing the brightness mapping model by using a bilateral structure evolution mode. Intelligent self-adaptive control of the brightness of the display screen of the intelligent storage cabinet can be realized, and the energy consumption is reduced while the visual experience of a user is improved.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent storage cabinet control, and in particular to a display screen brightness control method for an intelligent storage cabinet. Background Art

[0002] With the widespread application of smart storage cabinets in express delivery, supermarkets, libraries and other scenarios, the intelligent control of display brightness has become increasingly prominent. Reasonable display brightness control is not only related to the user's operating experience, but also directly affects the energy consumption of the equipment.

[0003] At present, common display brightness control methods mainly include fixed time period adjustment and single light sensor feedback adjustment. Among them, the fixed time period adjustment method is to pre-set the display brightness value for different time periods and make corresponding adjustments according to the current time; the single light sensor feedback adjustment method is to collect the ambient light intensity and determine the display brightness according to the preset mapping relationship.

[0004] The more advanced brightness control technology uses multiple light sensors to collect ambient light data and combines machine learning algorithms to establish a mapping relationship between ambient light and display brightness. This technology builds a brightness prediction model through training data and dynamically adjusts the display brightness based on real-time collected environmental data.

[0005] However, the existing technology still has the following problems: first, the existing brightness mapping model lacks a dynamic optimization mechanism and is difficult to adapt to dynamic changes in the environment and user needs; second, energy consumption constraints are not fully considered during the brightness adjustment process, which easily leads to energy waste; third, a single model structure is difficult to balance real-time and optimization effects, affecting the control effect. Summary of the invention

[0006] In view of this, the present application provides a display screen brightness control method for a smart storage cabinet, which solves the problems in the prior art that the brightness mapping model lacks a dynamic optimization mechanism, does not fully consider energy consumption constraints, and a single model structure is difficult to balance real-time and optimization effects.

[0007] The embodiment of the present application provides a display screen brightness control method for a smart storage cabinet, comprising: Acquire multiple light sensors to collect ambient light data and time data of the intelligent storage cabinet, pre-process the ambient light data and time data, and generate an environmental feature vector; Based on the environmental feature vector, a brightness mapping model is constructed using a neuroevolutionary algorithm to obtain an initial brightness value; Based on the initial brightness value, a constraint optimization method based on an Ising machine is used to perform brightness optimization to generate an optimized brightness value and optimization parameters; Using the optimized brightness value and optimization parameters, dynamically optimizing the brightness mapping model using a bilateral structural evolution mode to obtain optimized model parameters; wherein the optimized model parameters are used to update the brightness mapping model; The brightness of the display screen of the smart storage cabinet is controlled using the optimized brightness mapping model.

[0008] In some embodiments of the present application, obtaining multiple light sensors to collect ambient light data and time data of the smart storage cabinet includes: Based on a light sensor array consisting of a main sensor and a plurality of auxiliary sensors, collecting ambient light intensity data at a first preset frequency to generate an original light data set; Based on the current time information, a day is divided into multiple preset time periods, and the current season information is determined to generate a time data set.

[0009] In some embodiments of the present application, preprocessing the ambient light data includes: Based on the original illumination data set, using a statistical outlier detection algorithm to identify abnormal data and generate cleaned illumination data; Based on the cleaned illumination data, a sliding average algorithm of a first preset time window is used to perform smoothing processing to obtain the preprocessed illumination data.

[0010] In some embodiments of the present application, generating the environment feature vector includes: Performing minimum-maximum normalization processing on the preprocessed illumination data to obtain standardized illumination data; Based on the standardized illumination data and the time data set, the environmental feature vector is generated by a feature fusion algorithm.

[0011] In some embodiments of the present application, based on the environmental feature vector, a brightness mapping model is constructed using a neuroevolutionary algorithm to obtain an initial brightness value, including: Constructing a neural network structure with a preset number of layers based on the dimension of the environmental feature vector, initializing the network weights using an initialization method, and generating an initial network model; Generate a first preset number of initial populations, where the mutation probability is the first preset probability and the crossover probability is the second preset probability, to obtain a set of candidate network models; The fitness value of each candidate network model is calculated based on historical data, and the models with a third preset ratio whose fitness is greater than a preset threshold are retained to enter the next generation. After a second preset number of generations of evolution, a brightness mapping model is obtained, and the initial brightness value is output using the brightness mapping model.

[0012] In some embodiments of the present application, based on the initial brightness value, a constraint optimization method based on an Ising machine is used to perform brightness optimization, including: The brightness adjustment range is set to a first preset range, the energy consumption is limited to a second preset ratio of the maximum power, and a constraint satisfaction function is constructed based on the initial brightness value to obtain a constraint model; An Ising energy model is constructed based on the constraint model, and an energy function is designed, wherein the minimum value of the energy function corresponds to an optimal brightness adjustment scheme; An optimized brightness value satisfying the constraint condition is obtained based on the Ising energy model.

[0013] In some embodiments of the present application, obtaining an optimized brightness value satisfying a constraint condition based on the Ising energy model includes: The Ising energy model is solved by using a simulated annealing algorithm, wherein the initial temperature is a third preset value and the temperature reduction coefficient is a fourth preset value; A third preset number of state conversion attempts are performed at each temperature to obtain the optimized brightness value.

[0014] In some embodiments of the present application, a simulated annealing algorithm is used to solve the Ising energy model, including: Based on the energy function of the Ising energy model, the brightness value is encoded in a binary encoding manner to generate an initial state code; Based on the initial state encoding, generating candidate states by a single spin flip method, calculating the energy difference, and obtaining the state transition probability; Using the state transition probability and the current temperature parameter, the Metropolis criterion is used to determine whether to accept the candidate state, so as to generate the next state code; Based on the next state code, detecting the energy value stability and the temperature convergence condition, and if the temperature convergence condition is not met, updating the temperature parameter to generate an updated temperature parameter; Based on the updated temperature parameter, dynamically adjust the acceptance threshold of the Metropolis criterion, update the state transition strategy parameter, and generate the optimized state transition parameter; Performing a third preset number of state conversion attempts at each temperature to obtain the optimized brightness value includes: Based on the optimized state transfer parameters, a third preset number of state transitions are performed to obtain an optimal state code, which is decoded to generate the optimized brightness value.

[0015] In some embodiments of the present application, the brightness mapping model is dynamically optimized using the optimized brightness value and optimization parameters in a bilateral structural evolution mode, including: Collect the optimized brightness values ​​and historical optimization process data, extract the system optimization trajectory through a time series analysis method, and obtain a structural evolution model; Based on the structural evolution pattern, adding and deleting nodes, reconstructing connections and adjusting parameters of the network structure of the brightness mapping model are performed to generate an architecture optimization strategy; The brightness mapping model is updated based on the architecture optimization strategy.

[0016] In some embodiments of the present application, updating the brightness mapping model based on the architecture optimization strategy includes: Applying the architecture optimization strategy to the brightness mapping model, performing structural reorganization and parameter fine-tuning, and obtaining an updated network model; The performance of the updated network model is evaluated based on the validation data set, and the optimized model parameters are generated after confirming the performance improvement.

[0017] The present application also provides a computer device, the computer device comprising: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned method for controlling the brightness of the display screen of the smart storage cabinet.

[0018] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned method for controlling the brightness of a display screen for a smart storage cabinet.

[0019] An embodiment of the present application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the above-mentioned display screen brightness control method for a smart storage cabinet.

[0020] This application has the following technical effects: 1. By using a sensor array composed of multiple light sensors to collect ambient light data and combining it with time data for feature fusion, the accuracy and robustness of environmental feature extraction are improved; 2. The brightness mapping model was constructed using the neuroevolution algorithm, and the network structure and parameters were optimized through population evolution, which improved the adaptability and generalization ability of the model; 3. The constrained optimization method based on the Ising machine is used to optimize brightness and introduce energy consumption constraints, thus achieving optimal control of energy consumption while ensuring display effects; 4. The brightness mapping model is dynamically optimized through the bilateral structural evolution mode, which realizes the adaptive adjustment of the model structure and parameters and improves the long-term operation performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to make the technical solution of the present application clearer, the following is described in detail with reference to the accompanying drawings, wherein: Figure 1 A schematic diagram of a flow chart of a display screen brightness control method provided in an embodiment of the present application; Figure 2 This is a schematic diagram of the process of light data collection and preprocessing in the embodiment of the present application; Figure 3 A schematic diagram of the process of constructing a brightness mapping model based on a neuroevolutionary algorithm in an embodiment of the present application; Figure 4 This is a schematic diagram of the brightness optimization process based on the Ising machine in the embodiment of the present application; Figure 5 A schematic diagram of a process of solving the Ising energy model using a simulated annealing algorithm in an embodiment of the present application; Figure 6 A schematic diagram of the computer device structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. In addition, the technical features involved in each embodiment of the present application described below can be combined with each other as long as they do not conflict with each other.

[0023] like Figure 1 As shown, an embodiment of the present application provides a display screen brightness control method for a smart storage cabinet, comprising: S1: Acquire multiple light sensors to collect ambient light data and time data of the smart storage cabinet, pre-process the ambient light data and time data, and generate an environmental feature vector; like Figure 2 As shown, obtaining multiple light sensors to collect ambient light data and time data of the smart storage cabinet includes: S1.1 collects ambient light intensity data at a first preset frequency based on a light sensor array consisting of a main sensor and a plurality of auxiliary sensors to generate an original light data set; The technical solution of the present application is described in detail below in conjunction with specific embodiments.

[0024] In this embodiment, the light sensor array includes a main sensor and multiple auxiliary sensors located around the display screen of the smart storage cabinet. The main sensor is usually arranged directly above the display screen to collect the light intensity in the main visual area of ​​the user; the auxiliary sensors are distributed around the display screen to collect the spatial distribution information of the ambient light. Specifically, the first preset frequency can be set to 1Hz, that is, light data is collected once per second. This sampling frequency can ensure timely response to changes in ambient light without generating excessive data processing burden.

[0025] S1.2 divides a day into a plurality of preset time periods based on the current time information, determines the current season information, and generates a time data set.

[0026] Regarding the processing of time data, this embodiment divides a day into six preset time periods: early morning (0:00-5:59), morning (6:00-8:59), morning (9:00-11:59), afternoon (12:00-16:59), evening (17:00-19:59) and night (20:00-23:59). This division method takes into account the changing pattern of natural light during the day and the user's habit of using smart storage cabinets. At the same time, the system will also determine the season information based on the current date, and divide the whole year into spring (March-May), summer (June-August), autumn (September-November) and winter (December-February) to better adapt to the lighting characteristics of different seasons.

[0027] The preprocessing of the ambient light data includes: Based on the original illumination data set, an abnormal data is identified by using a statistical outlier detection algorithm to generate cleaned illumination data; based on the cleaned illumination data, a sliding average algorithm of a first preset time window is used for smoothing to obtain the preprocessed illumination data.

[0028] Specifically, the embodiment of the present application adopts a statistically based outlier detection algorithm to identify abnormal data. Specifically, the system uses an improved Z-score method for outlier detection, and the calculation formula is: Z = |x - μ| / σ Where x is the illumination data to be detected, μ is the mean value in the sliding time window, and σ is the standard deviation. When the Z value exceeds the preset threshold (usually set to 3), the data point is determined to be an outlier. For detected outliers, the system replaces them with the average value of the adjacent valid data points to generate cleaned illumination data.

[0029] Next, the cleaned illumination data is smoothed. This embodiment uses a sliding average algorithm of a first preset time window, and the window size is set to 30 sampling points (i.e., 30 seconds). In each sliding window, the weighted average of all data points is calculated, where newer data points have larger weights. This processing method can effectively suppress short-term fluctuations in illumination data while maintaining sensitivity to changes in ambient illumination.

[0030] S1.3: Generating the environmental feature vector includes: Performing minimum-maximum normalization processing on the preprocessed illumination data to obtain standardized illumination data; Based on the standardized illumination data and the time data set, the environmental feature vector is generated by a feature fusion algorithm.

[0031] In S1.3, the preprocessed illumination data is subjected to minimum-maximum normalization. The normalization formula is: x_norm = (x - x_min) / (x_max - x_min) Among them, x is the original data, x_min and x_max are the minimum and maximum values ​​in the data set respectively. After this processing, all illumination data are mapped to the interval [0,1], which is convenient for subsequent feature fusion.

[0032] The feature fusion process uses a multimodal feature fusion algorithm to fuse the standardized illumination data and time data sets. Specifically, for illumination data, its intensity value, rate of change, and spatial distribution characteristics are extracted; for time data, the time period and season information are converted into one-hot encoding. Then, a fixed-dimensional environmental feature vector is generated by feature concatenation and weighted fusion. Among them, the weight of the illumination feature is 0.7, and the weight of the time feature is 0.3. This weight ratio shows good results in actual tests.

[0033] S2: Based on the environmental feature vector, a brightness mapping model is constructed using a neuroevolutionary algorithm to obtain an initial brightness value; S2.1: constructing a neural network structure with a preset number of layers based on the dimension of the environmental feature vector, initializing the network weights using an initialization method, and generating an initial network model; like Figure 3As shown, first, a neural network structure is constructed based on the dimension of the environmental feature vector. Considering that the environmental feature vector contains multi-dimensional features such as light intensity, rate of change, spatial distribution, and time coding, this embodiment adopts a four-layer feedforward neural network structure: the number of nodes in the input layer is the same as the dimension of the environmental feature vector (set to N), the two hidden layers contain 2N and N nodes respectively, and the output layer is 1 node, corresponding to the brightness value of the display screen. The network uses ReLU (Rectified Linear Unit) as the activation function to provide nonlinear modeling capabilities. For the initialization of network weights, the improved He initialization method is used, and the initial weight values ​​obey a normal distribution with a mean of 0 and a standard deviation of sqrt(2 / n), where n is the input dimension of the layer. This initialization method can effectively avoid the gradient vanishing problem in deep networks.

[0034] S2.2: Generate a first preset number of initial populations, where the mutation probability is the first preset probability and the crossover probability is the second preset probability, to obtain a set of candidate network models; When generating the initial population, the first preset number is set to 50, that is, the population size of 50 candidate network models is maintained. Each candidate model has the same network structure, but different initial weights. Mutation operations include weight mutation and structural mutation: weight mutation is to add random perturbations that obey the normal distribution N(0, 0.1) to the selected weight parameters; structural mutation may add or delete nodes in the hidden layer and change the inter-layer connection. The first preset probability (mutation probability) is set to 0.2, and the second preset probability (crossover probability) is set to 0.8. The crossover operation adopts an improved uniform crossover method, and the weight parameters at the corresponding positions have an equal probability of coming from any individual in the parent generation.

[0035] S2.3: Calculate the fitness value of each candidate network model based on historical data, retain the models with a third preset ratio whose fitness is greater than the preset threshold to enter the next generation, and after a second preset number of generations of evolution, obtain a brightness mapping model, and use the brightness mapping model to output the initial brightness value.

[0036] Fitness calculation is a key link in the evolution process. In this embodiment, the fitness function comprehensively considers three aspects: display effect, energy consumption control and response time. The specific form is: Fitness = w1 (1 - MSE) + w2 (1 - P / Pmax) + w3 (1 - T / Tmax) Among them, MSE is the mean square error between the model prediction value and the historical optimal brightness value, P is the power consumption corresponding to the predicted brightness, Pmax is the maximum power limit, T is the model inference time, and Tmax is the maximum allowed response time. The weight coefficients w1, w2, and w3 are set to 0.5, 0.3, and 0.2, respectively, reflecting the priority of the display effect. The third preset ratio is set to 0.4, that is, the top 40% of individuals in fitness are retained in each generation. The second preset number (evolutionary generations) is set to 100 generations, which shows good convergence in the experiment.

[0037] In order to speed up the evolution process, this embodiment also adopts an elite retention strategy, which directly retains the best three individuals in each generation to the next generation. At the same time, an adaptive mutation rate mechanism is introduced: when the population diversity decreases, the mutation probability is appropriately increased to avoid falling into the local optimum. Specifically, the population diversity is measured by the standard deviation of the fitness between individuals. When the standard deviation is less than the preset threshold (set to 0.1), the mutation probability increases by 50%.

[0038] After evolutionary optimization, the network model with the highest fitness is selected as the final brightness mapping model. The model receives the environmental feature vector as input and outputs a standardized brightness value in the range of [0,1]. Finally, it is mapped to the actual brightness range (usually 0-255) through denormalization to obtain the initial brightness value. The entire evolution process is accelerated by parallel computing, and an early stopping mechanism is adopted: if the optimal fitness does not improve significantly for 10 consecutive generations (the improvement is less than 0.001), the evolution process is terminated early.

[0039] S3: Based on the initial brightness value, a constraint optimization method based on an Ising machine is used to perform brightness optimization to generate an optimized brightness value and optimization parameters; like Figure 4 As shown, based on the initial brightness value, a constraint optimization method based on an Ising machine is used to perform brightness optimization, including: S3.1: setting the brightness adjustment range to a first preset range, limiting the energy consumption to a second preset ratio of the maximum power, constructing a constraint satisfaction function based on the initial brightness value, and obtaining a constraint model; First, this embodiment defines the constraints of brightness adjustment. The brightness adjustment range (first preset range) is set to [Bmin, Bmax], where Bmin is the minimum allowed brightness value (set to 20 to ensure basic visibility) and Bmax is the maximum brightness value (set to 255). Considering the need for energy consumption control, the energy consumption limit is set to the second preset ratio of the maximum power (set to 80%). In practical applications, the power consumption of the display screen is approximately quadratic with the brightness value, which can be expressed as: P(b) = k b^2, where k is the device-related scaling factor and b is the brightness value.

[0040] S3.2: constructing an Ising energy model based on the constraint model, and designing an energy function, wherein the minimum value of the energy function corresponds to an optimal brightness adjustment solution; Based on the above constraints, a constraint satisfaction function is constructed. This function comprehensively considers three aspects of constraints: brightness range constraint, energy consumption constraint and deviation constraint from initial brightness. The constraint satisfaction function is defined as follows: S(b) = w1 R(b) + w2 E(b) + w3 D(b) Among them, R(b) represents the satisfaction of the brightness range constraint: R(b) = 1, when b∈[Bmin, Bmax] R(b) = exp(-α|b-Bbound|), when b is out of range Here, Bbound is the nearest boundary value and α is the penalty factor (set to 0.1).

[0041] E(b) represents the satisfaction of energy consumption constraint: E(b) = 1, when P(b) ≤ Pmax 80% E(b) = exp(-β(P(b) / (Pmax 0.8) - 1)), when P(b) > Pmax 80% Among them, β is the penalty factor for exceeding the energy consumption limit (set to 0.2).

[0042] D(b) represents the deviation constraint from the initial brightness b0: D(b) = exp(-γ|b-b0| / b0) Among them, γ is the deviation penalty factor (set to 0.15). The weight coefficients w1, w2, and w3 are set to 0.4, 0.4, and 0.2 respectively, reflecting the importance of brightness range and energy consumption control.

[0043] Next, the constrained optimization problem is converted into the Ising energy model. First, the brightness value b is quantized into N-bit binary representation (N is usually 8), and each bit corresponds to a spin variable si∈{-1,1}. The Ising energy function is designed as: H = -J∑(si sj) - h∑si + λ(1-S(b)) Among them, J represents the coupling strength between spins (set to 1.0), h is the external field strength (set to 0.5), and λ is the constraint penalty factor (set to 2.0). The first term represents the interaction between spins, which promotes the consistency of adjacent bits; the second term introduces external field bias to avoid falling into local optimality; the third term converts constraint satisfaction into an energy term. The greater the degree of constraint violation, the higher the energy.

[0044] The Ising energy function has the following characteristics: the minimum energy value corresponds to the state with the maximum constraint satisfaction, that is, the optimal brightness adjustment scheme; the local structure of the energy function reflects the physical constraints of brightness adjustment, which is conducive to the convergence of the optimization algorithm; by adjusting the parameters J, h and λ, the importance of different constraints can be balanced.

[0045] S3.3: Obtaining an optimized brightness value that satisfies constraint conditions based on the Ising energy model.

[0046] By minimizing the above Ising energy function, the optimized brightness value that satisfies the constraints can be obtained. The specific solution process will be implemented by the simulated annealing algorithm in the subsequent steps. This constrained optimization method based on the Ising model has the following advantages: it can naturally handle discrete optimization problems, it can flexibly integrate multiple constraints, and the optimization process has good physical interpretation.

[0047] Wherein, obtaining an optimized brightness value satisfying the constraint condition based on the Ising energy model includes: A1. Using a simulated annealing algorithm to solve the Ising energy model, wherein the initial temperature is a third preset value and the cooling coefficient is a fourth preset value; A2. Perform a third preset number of state conversion attempts at each temperature to obtain the optimized brightness value.

[0048] The specific implementation process of solving the Ising energy model using the simulated annealing algorithm is described in detail below.

[0049] This embodiment uses an improved simulated annealing algorithm to solve the Ising energy model. The algorithm simulates the annealing process of the physical system, allowing the system to fully explore the state space at high temperatures, and gradually converge to the state with minimum energy as the temperature decreases. The initial temperature (the third preset value) is set to T0 = 100, which is high enough to ensure that the system has enough energy to cross the local minimum in the initial stage. The cooling coefficient (the fourth preset value) is set to α = 0.95, that is, the temperature is reduced by 5% after each iteration. This slow cooling strategy helps the system to fully explore the state space and avoid falling into the local optimal solution too early.

[0050] At each temperature, the system needs to make a sufficient number of state transition attempts to reach a quasi-equilibrium state. In this embodiment, the third preset number is set to 1000 times, that is, 1000 state transition attempts are made at each temperature point. This value is obtained through experimental optimization and can achieve a good balance between computing efficiency and optimization effect. Specifically, the state transition process is as follows: First, for each state transition attempt, the algorithm randomly selects one or more spin variables to flip. In order to improve the search efficiency, this embodiment adopts an adaptive multi-spin flip strategy: in the high temperature stage (T > 0.7T0), up to 4 spins may be flipped at the same time each time to speed up the exploration of the state space; in the medium temperature stage (0.7T0 ≥ T > 0.3T0), up to 2 spins are flipped at the same time; in the low temperature stage (T ≤ 0.3T0), only 1 spin is flipped at a time to achieve fine adjustment.

[0051] Then, the energy difference ΔE before and after the state transition is calculated. If the energy difference is negative, that is, the energy of the new state is lower, the transition is accepted directly; if the energy difference is positive, the transition is accepted with probability P = exp(-ΔE / T). This acceptance mechanism based on the Metropolis criterion allows the system to accept energy-increasing transitions with a certain probability, thereby avoiding falling into a local optimal solution.

[0052] In order to improve the convergence efficiency of the algorithm, this embodiment also introduces the following optimization strategies: 1. Historical optimal solution record: maintain a historical optimal solution pool to record the optimal solutions encountered during the search process and their corresponding energy values; 2. Temperature adaptive adjustment: If 100 consecutive state transitions are not accepted, increase the current temperature (multiply by 1.1) to increase the probability of accepting a new solution; 3. Restart mechanism: If it is found that the system may fall into a local optimum (no better solution is obtained after 500 consecutive conversions), it will be reset to the historical optimal solution with a probability of 0.2 and reinitialized to a random state with a probability of 0.8.

[0053] The termination condition of the algorithm is set as follows: when the temperature is lower than the threshold Tmin (set to 0.01) or no better solution is obtained after 10 consecutive temperature updates, the annealing process ends. Finally, the binary code corresponding to the optimal solution found in the search process is converted into a brightness value, that is, the optimized brightness value that meets the constraints is obtained.

[0054] Through the improved simulated annealing algorithm, the present embodiment can efficiently solve the Ising energy model, while ensuring the quality of the solution, with a faster convergence speed. Experiments show that when the algorithm processes the 8-bit quantized brightness value optimization problem, the average convergence time does not exceed 100ms, meeting the requirements of real-time control.

[0055] In one embodiment, if Figure 5 As shown, the Ising energy model is solved by using a simulated annealing algorithm, including: B1. Based on the energy function of the Ising energy model, the brightness value is encoded using a binary encoding method to generate an initial state code; like Figure 5 As shown, the specific steps of using simulated annealing algorithm to solve the Ising energy model are as follows.

[0056] First, the brightness value needs to be binary encoded. This embodiment uses an 8-bit binary encoding method to convert the brightness value (range 0-255) into a corresponding binary sequence. For the Ising model, it is necessary to map binary 0 and 1 to spin states -1 and +1. For example, the binary representation of the brightness value 128 is 10000000, and the corresponding initial state is encoded as [+1,-1,-1,-1,-1,-1,-1,-1]. This encoding method ensures consistency with the spin representation of the Ising model while maintaining the physical meaning of the brightness value.

[0057] B2 generates candidate states based on the initial state encoding by a single spin flip method, calculates the energy difference, and obtains the state transition probability; When generating candidate states, this embodiment adopts a single spin flip strategy. Specifically, in each state transition attempt, a spin at one position is randomly selected for flipping (i.e., from +1 to -1 or from -1 to +1). For example, if the current state is encoded as [+1,-1,-1,-1,-1,-1,-1,-1], the third position is randomly selected for flipping, and the generated candidate state is [+1,-1,+1,-1,-1,-1,-1,-1]. Then, the energy difference ΔE before and after the state transition is calculated based on the Ising energy function. The calculation of the energy difference needs to take into account the changes in the interaction terms, external field terms, and constraint terms between the spins.

[0058] B3 uses the state transition probability and the current temperature parameter and the Metropolis criterion to determine whether to accept the candidate state so as to generate the next state code; For the calculated energy difference, the system uses the Metropolis criterion to determine whether to accept the candidate state. The specific judgment process is as follows: if ΔE ≤ 0, directly accept the candidate state; if ΔE > 0, accept the candidate state with probability P = exp(-ΔE / T), where T is the current temperature parameter. This probabilistic acceptance mechanism makes the system more likely to accept a state with increased energy at high temperatures, and tend to accept a state with decreased energy at low temperatures.

[0059] After each state transition, the system needs to detect the stability of the energy value and the convergence of the temperature. This embodiment uses a sliding window method to monitor the fluctuation of the energy value: in the last 100 state transitions, if the standard deviation of the energy value is less than a preset threshold (for example, 0.01), it is considered that the system has reached a quasi-equilibrium state at the current temperature. Temperature convergence conditions include: the current temperature is lower than the minimum temperature threshold (0.01), the energy value has not improved significantly after 5 consecutive temperature updates (the improvement is less than 0.1%), or the maximum number of iterations (10,000 times) has been reached.

[0060] B4 detects the energy value stability and the temperature convergence condition based on the next state code, and updates the temperature parameter if the temperature convergence condition is not met, to generate an updated temperature parameter; When the temperature convergence condition is not met, the system updates the temperature parameters. The temperature update adopts an improved exponential cooling strategy: T_new = α T_current, where the cooling coefficient α will be dynamically adjusted according to the system state. Specifically, when the energy value fluctuates greatly (the standard deviation is greater than 0.1), α takes a larger value (0.95) to maintain a higher exploration capability; when the energy value tends to be stable, α takes a smaller value (0.90) to accelerate convergence.

[0061] B5 dynamically adjusts the acceptance threshold of the Metropolis criterion based on the updated temperature parameter, updates the state transition strategy parameter, and generates an optimized state transition parameter; Based on the updated temperature parameters, the system dynamically adjusts the acceptance threshold of the Metropolis criterion. The adjustment formula for the acceptance threshold is: threshold = min(0.9, max(0.1, T / T0)), where T0 is the initial temperature. This dynamic threshold mechanism ensures the balance between exploration and utilization of the system during the annealing process: at high temperatures, more energy-increasing states are accepted to explore the solution space, and at low temperatures, the energy-reducing states are mainly accepted to achieve local optimization.

[0062] Performing a third preset number of state conversion attempts at each temperature to obtain the optimized brightness value includes: B6 performs a third preset number of state transitions based on the optimized state transfer parameters to obtain an optimal state code, and generates the optimized brightness value after decoding.

[0063] Finally, a third preset number (1000) of state transition attempts are performed at each temperature. The system records the optimal state encoding encountered during these transitions and decodes it into the corresponding brightness value. The decoding process includes: first mapping the spin state [+1,-1] back to binary [1,0], and then converting the binary sequence to decimal numbers, that is, obtaining the optimized brightness value. In order to ensure the validity of the decoding result, the system will perform a range check on the decoded brightness value to ensure that it falls within the preset brightness adjustment range.

[0064] S4: using the optimized brightness value and optimization parameters, dynamically optimizing the brightness mapping model using a bilateral structural evolution mode to obtain optimized model parameters; wherein the optimized model parameters are used to update the brightness mapping model; The optimized brightness value and optimization parameters are used to dynamically optimize the brightness mapping model using a bilateral structural evolution mode, including: S4.1: collecting the optimized brightness value and historical optimization process data, extracting the system optimization trajectory through a time series analysis method, and obtaining a structural evolution model; The specific implementation process of dynamically optimizing the brightness mapping model using the bilateral structure evolution mode is described in detail below.

[0065] This embodiment first establishes an optimization history database to store key data during the operation of the system. Specifically, for each brightness optimization process, the following data are recorded: the original environment feature vector, the initial brightness value output by the neural network, the brightness value after Ising optimization, the energy change trajectory during the optimization process, the model prediction error, etc. The system uses a sliding time window method to organize these historical data, and the window size is set to the most recent 1000 optimization records to ensure that the data can reflect the long-term behavior pattern of the system and respond to environmental changes in a timely manner.

[0066] In the timing analysis stage, this embodiment uses a multi-scale analysis method to extract the characteristics of the system optimization trajectory. First, on a small time scale (within 10 minutes), the deviation pattern between the model prediction value and the optimized brightness value is analyzed, and the mean, variance, and autocorrelation characteristics of the deviation are calculated. Secondly, on a medium time scale (within 24 hours), the periodic pattern and trend characteristics of brightness adjustment are identified, and the wavelet transform method is used to decompose different frequency components. Finally, on a large time scale (more than 7 days), the overall performance change trend of the model is evaluated, including indicators such as prediction accuracy, energy efficiency, and response time.

[0067] Based on the above time series analysis results, the system constructs a structural evolution model. The evolution model contains three key dimensions: prediction accuracy, model complexity and computational efficiency. Specifically: 1. When continuous prediction deviation is detected (for example, the prediction error exceeds 10% for 100 consecutive times), the system determines that the current network structure has insufficient expressive power and needs to increase network capacity; 2. If it is found that the output of some nodes has a low contribution to the final prediction result (for example, the absolute value of the weight is less than 0.01) for more than 1,000 optimization cycles, consider deleting these redundant nodes; 3. By analyzing the activation patterns of nodes at different layers, we can identify groups of nodes that are frequently co-activated, which can be used as a basis for adjusting the network connection structure.

[0068] S4.2: Based on the structure evolution pattern, adding and deleting nodes, reconstructing connections and adjusting parameters of the network structure of the brightness mapping model are performed to generate an architecture optimization strategy; Based on the structural evolution pattern, the system generates specific architecture optimization strategies. The node addition and deletion strategy adopts a progressive method: at most 10% of the number of nodes in the current layer is added or deleted each time to maintain the stability of the network structure. The connection reconstruction adopts the attention mechanism to dynamically assign importance weights to different input features. Specifically, the system maintains a feature importance matrix, and the matrix element values ​​are dynamically updated according to the contribution of the feature to the prediction accuracy. The update formula is: W_new = β W_old + (1-β) ΔAccuracy Among them, β is the smoothing factor (set to 0.9), and ΔAccuracy is the marginal contribution of this feature to the prediction accuracy.

[0069] The parameter adjustment adopts a two-way propagation mechanism: in forward propagation, the weight distribution of each layer is adjusted according to the feature importance matrix; in reverse propagation, the brightness value after Ising optimization is used as a supervision signal to update the network parameters. In order to prevent excessive adjustment, the system sets a dynamic learning rate for parameter update: η = η_base exp(-λ performance_score) Among them, η_base is the basic learning rate (set to 0.01), λ is the decay factor (set to 0.1), and performance_score is the comprehensive performance score of the current model.

[0070] This two-sided structural evolution model is adaptive and stable: the expressiveness of the model is improved through positive structural optimization, and the stability of the optimization is ensured through reverse parameter adjustment. Experiments show that this method can enable the brightness mapping model to maintain good prediction performance in the long-term operation process, while effectively controlling the computational complexity of the model.

[0071] S4.3: Update the brightness mapping model based on the architecture optimization strategy.

[0072] In S4.3, updating the brightness mapping model based on the architecture optimization strategy includes: C1. applying the architecture optimization strategy to the brightness mapping model, performing structural reorganization and parameter fine-tuning, and obtaining an updated network model; C2. Evaluate the performance of the updated network model based on the validation data set, and generate the optimized model parameters after confirming the performance improvement.

[0073] The specific implementation process and model evaluation method of the architecture optimization strategy are described in detail below.

[0074] When performing structural reorganization, the system applies architectural optimization strategies in a hierarchical and progressive manner. First, the connections between the input layer and the first hidden layer are reorganized, the connection weights are adjusted based on the feature importance matrix, and the connections with importance below the threshold (set to 0.05) are pruned. Then, the nodes of the first hidden layer are evaluated and the activity index of each node is calculated: Activity_score = α activation_frequency + (1-α) weight_magnitude Among them, activation_frequency represents the activation frequency of the node in the last 1000 inferences, weight_magnitude represents the L1 norm of the node's related weight, and α is the balance factor (set to 0.6). Based on the activity index, the system retains the 80% of nodes with the highest scores and adds new nodes in high-load areas (areas with activity greater than 0.8).

[0075] Parameter fine-tuning uses an improved elastic weight merging method. For nodes that need to be merged, their parameters are calculated as follows: W_merged = (W1 S1 + W2 S2) / (S1 + S2) Among them, W1 and W2 are the weights of the nodes to be merged, and S1 and S2 are the corresponding importance scores. For the newly added nodes, their initial parameters are obtained by interpolating the parameters of the adjacent nodes, and a small random perturbation (the standard deviation is set to 0.01) is added to break the symmetry. In order to maintain the stability of network functions, the parameter fine-tuning process adopts a progressive update strategy, and the parameter value is adjusted by a maximum of 20% each time.

[0076] In the model evaluation phase, the system builds a multi-dimensional verification indicator system. The verification data set contains three types of data: 1. Historical optimization data: environmental characteristics and optimized brightness value pairs for the last 7 days; 2. Typical scene data: standard brightness values ​​under 10 pre-defined typical lighting environments; 3. Boundary condition data: including test cases under extreme lighting conditions and rapidly changing scenes.

[0077] Performance evaluation adopts a weighted comprehensive scoring method: Score = w1 Accuracy + w2 Stability + w3 Efficiency Among them, Accuracy measures the prediction accuracy (expressed as root mean square error), Stability evaluates the stability of the model in different scenarios (expressed as prediction variance), and Efficiency measures the computational efficiency (expressed as single inference time). The weight coefficients w1, w2, and w3 are set to 0.5, 0.3, and 0.2 respectively.

[0078] The system sets up a strict performance improvement confirmation mechanism: the comprehensive score of the new model must be at least 3% higher than the current model, and no single indicator can drop by more than 5%. At the same time, a stability observation window is introduced: the model performance is monitored during 100 consecutive inferences, and the optimized model parameters are generated only when the performance improvement is maintained throughout the observation window.

[0079] If the model passes the performance evaluation, the system will generate a model description file containing the complete network structure and weight parameters. The file uses a hierarchical storage structure, including network topology information, node parameters, connection weights, etc., to facilitate subsequent model deployment and version control. If the model fails the performance evaluation, the system will roll back to the last stable version and record the failure reason in the log to provide a reference for subsequent optimization.

[0080] S5: Using the optimized brightness mapping model to control the brightness of the display screen of the smart storage cabinet.

[0081] This embodiment illustrates the implementation process of brightness control by using a specific example. Assuming that in a smart storage cabinet scenario in a shopping mall, consider the following typical situation: Example 1: Direct sunlight scene: At 10 am, sunlight shines directly into the smart storage cabinet area through the mall glass. At this time: 1) The data collected by the light sensor array show that the light intensity of the main sensor is 50,000 lux, and the light intensities of the surrounding auxiliary sensors are 45,000 lux, 48,000 lux, 42,000 lux, and 46,000 lux respectively; 2) The time feature shows that it is currently in the morning and the season is summer; 3) After feature fusion, the environmental feature vector reflects the characteristics of strong and relatively uniform lighting; 4) The optimized brightness mapping model outputs a display brightness value of 240 (range 0-255), ensuring visibility of displayed content in strong lighting environments.

[0082] Example 2: Indoor low-light scene: At 21:00, the mall is about to close and the area relies mainly on artificial lighting. 1) The data collected by the light sensor array shows that the light intensity of the main sensor is 200 lux, and the light intensities of the surrounding auxiliary sensors are 180 lux, 210 lux, 190 lux and 200 lux respectively; 2) The time feature shows that it is currently night time and the season is winter; 3) After feature fusion, the environmental feature vector reflects the weak light and relatively stable features; 4) The optimized brightness mapping model outputs a display brightness value of 60, which reduces energy consumption while ensuring that the displayed content is clearly visible.

[0083] Example 3: Scene with rapid lighting changes: At 16:00 in the afternoon, the light intensity changes frequently due to changes in cloud cover. At this time: 1) The light sensor collects the main sensor light intensity fluctuation from 5000 lux to 20000 lux within 10 consecutive seconds; 2) The system uses a sliding average algorithm to smooth the light data and suppress short-term fluctuations; 3) The brightness mapping model combines historical optimization experience to output a relatively stable brightness value sequence to avoid frequent changes in display brightness; 4) The final display brightness is smoothly adjusted within the range of 140-160, which not only adapts to environmental changes but also maintains a good visual experience.

[0084] In actual applications, the system samples and evaluates the environment once a second, but in order to avoid frequent adjustments to the brightness, a minimum adjustment interval (10 seconds) and a minimum adjustment amplitude (brightness changes must exceed 5 units) are set. At the same time, the system also records the time distribution characteristics of users using the smart storage cabinet, and maintains a more active brightness adjustment strategy during high-frequency use periods (such as rush hour) to provide a better user experience.

[0085] From the above examples, it can be seen that the brightness control method provided by the present application can adaptively adjust the brightness of the display screen according to the environmental characteristics of different scenes, which not only ensures the visibility of the displayed content, but also realizes the optimization control of energy consumption. Especially in actual application scenarios with complex and changeable lighting conditions, this method shows good environmental adaptability and control stability.

[0086] The present application also provides a computer device, such as Figure 6 As shown, the computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the above-mentioned display screen brightness control method for the smart storage cabinet.

[0087] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned display screen brightness control method for a smart storage cabinet.

[0088] An embodiment of the present application also provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the above-mentioned display screen brightness control method for a smart storage cabinet are implemented.

[0089] Each module in the display screen brightness control system for the smart storage cabinet can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0090] In one embodiment, a computer device is also provided, which is the display screen brightness control system for the smart storage cabinet mentioned in the above method embodiment. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface.

[0091] Among them, the processor of the computer device is used to provide computing and control capabilities, and can be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but is not limited thereto. The processor may include one or more processors, for example, one or more central processing units (CPUs). When the processor is a CPU, the CPU may be a single-core CPU or a multi-core CPU. The processor may also include one or more dedicated processors, and the dedicated processor may include a GPU, an FPGA, etc., for accelerated processing. The processor is used to call the program code and data in the memory and execute the steps in the above method embodiment. For details, please refer to the description in the method embodiment, which will not be repeated here.

[0092] The memory of the computer device includes, but is not limited to, non-volatile storage media and internal memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium.

[0093] The input / output interface of the computer device is used to exchange information between the processor and external devices.

[0094] The communication interface of the computer device is used to communicate with an external terminal via a network connection.

[0095] When the computer program is executed by a processor, the above method is implemented.

[0096] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the division of each unit / module is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling, direct coupling, or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or units, which can be electrical, mechanical or other forms.

[0097] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0098] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a read-only memory (ROM), or a random access memory (RAM), or a magnetic medium, such as a floppy disk, a hard disk, a tape, a disk, or an optical medium, such as a digital versatile disc (DVD), or a semiconductor medium, such as a solid state disk (SSD), etc.

[0099] The system of the embodiments of the present disclosure can execute the method provided by the embodiments of the present disclosure, and the implementation principles are similar. The actions performed by each module in the system of each embodiment of the present disclosure correspond to the steps in the method of each embodiment of the present disclosure. For the detailed functional description of each module of the system, please refer to the description in the corresponding method shown in the previous text, which will not be repeated here.

[0100] The above description is only an optional implementation method for some implementation scenarios of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the scheme of the present disclosure, other similar implementation methods based on the technical ideas of the present disclosure are also within the protection scope of the embodiments of the present disclosure.

[0101] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A display screen brightness control method for a smart storage cabinet, characterized in that: include: Acquire multiple light sensors to collect ambient light data and time data of the intelligent storage cabinet, pre-process the ambient light data and time data, and generate an environmental feature vector; Based on the environmental feature vector, a brightness mapping model is constructed using a neuroevolutionary algorithm to obtain an initial brightness value; Based on the initial brightness value, a constraint optimization method based on an Ising machine is used to perform brightness optimization to generate an optimized brightness value and optimization parameters; Using the optimized brightness value and optimization parameters, dynamically optimizing the brightness mapping model using a bilateral structural evolution mode to obtain optimized model parameters; wherein the optimized model parameters are used to update the brightness mapping model; The brightness of the display screen of the smart storage cabinet is controlled using the optimized brightness mapping model.

2. The method according to claim 1, characterized in that Acquire multiple light sensors to collect ambient light data and time data of the smart storage cabinet, including: Based on a light sensor array consisting of a main sensor and a plurality of auxiliary sensors, collecting ambient light intensity data at a first preset frequency to generate an original light data set; Based on the current time information, a day is divided into multiple preset time periods, and the current season information is determined to generate a time data set.

3. The method according to claim 2, characterized in that Preprocessing the ambient light data includes: Based on the original illumination data set, using a statistical outlier detection algorithm to identify abnormal data and generate cleaned illumination data; Based on the cleaned illumination data, a sliding average algorithm of a first preset time window is used to perform smoothing processing to obtain the preprocessed illumination data.

4. The method according to claim 3, characterized in that Generating the environmental feature vector includes: Performing minimum-maximum normalization processing on the preprocessed illumination data to obtain standardized illumination data; Based on the standardized illumination data and the time data set, the environmental feature vector is generated by a feature fusion algorithm.

5. The method according to claim 1, characterized in that Based on the environmental feature vector, a brightness mapping model is constructed using a neuroevolutionary algorithm to obtain an initial brightness value, including: Constructing a neural network structure with a preset number of layers based on the dimension of the environmental feature vector, initializing the network weights using an initialization method, and generating an initial network model; Generate a first preset number of initial populations, where the mutation probability is the first preset probability and the crossover probability is the second preset probability, to obtain a set of candidate network models; The fitness value of each candidate network model is calculated based on historical data, and the models with a third preset ratio whose fitness is greater than a preset threshold are retained to enter the next generation. After a second preset number of generations of evolution, a brightness mapping model is obtained, and the initial brightness value is output using the brightness mapping model.

6. The method according to claim 1, characterized in that Based on the initial brightness value, brightness optimization is performed using a constraint optimization method based on an Ising machine, including: The brightness adjustment range is set to a first preset range, the energy consumption is limited to a second preset ratio of the maximum power, and a constraint satisfaction function is constructed based on the initial brightness value to obtain a constraint model; An Ising energy model is constructed based on the constraint model, and an energy function is designed, wherein the minimum value of the energy function corresponds to an optimal brightness adjustment scheme; An optimized brightness value satisfying the constraint condition is obtained based on the Ising energy model.

7. The method according to claim 6, characterized in that Obtaining an optimized brightness value that satisfies the constraint conditions based on the Ising energy model includes: The Ising energy model is solved by using a simulated annealing algorithm, wherein the initial temperature is a third preset value and the temperature reduction coefficient is a fourth preset value; A third preset number of state conversion attempts are performed at each temperature to obtain the optimized brightness value.

8. The method according to claim 7, characterized in that The Ising energy model is solved by using a simulated annealing algorithm, including: Based on the energy function of the Ising energy model, the brightness value is encoded in a binary encoding manner to generate an initial state code; Based on the initial state encoding, generating candidate states by a single spin flip method, calculating the energy difference, and obtaining the state transition probability; Using the state transition probability and the current temperature parameter, the Metropolis criterion is used to determine whether to accept the candidate state, so as to generate the next state code; Based on the next state code, detecting the energy value stability and the temperature convergence condition, and if the temperature convergence condition is not met, updating the temperature parameter to generate an updated temperature parameter; Based on the updated temperature parameter, dynamically adjust the acceptance threshold of the Metropolis criterion, update the state transition strategy parameter, and generate the optimized state transition parameter; Performing a third preset number of state conversion attempts at each temperature to obtain the optimized brightness value includes: Based on the optimized state transfer parameters, a third preset number of state transitions are performed to obtain an optimal state code, which is decoded to generate the optimized brightness value.

9. The method according to claim 1, characterized in that: Using the optimized brightness value and optimization parameters, the brightness mapping model is dynamically optimized using a bilateral structural evolution mode, including: Collect the optimized brightness values ​​and historical optimization process data, extract the system optimization trajectory through a time series analysis method, and obtain a structural evolution model; Based on the structural evolution pattern, adding and deleting nodes, reconstructing connections and adjusting parameters of the network structure of the brightness mapping model are performed to generate an architecture optimization strategy; The brightness mapping model is updated based on the architecture optimization strategy.

10. The method according to claim 9, characterized in that Updating the brightness mapping model based on the architecture optimization strategy includes: Applying the architecture optimization strategy to the brightness mapping model, performing structural reorganization and parameter fine-tuning, and obtaining an updated network model; The performance of the updated network model is evaluated based on the validation data set, and the optimized model parameters are generated after confirming the performance improvement.

Images