An imaging parameter optimization method for visual imaging

Abstract

This invention discloses an imaging parameter optimization method for visual imaging, relating to the field of electronic visual imaging technology. The method includes receiving parameter configurations output by an optimization algorithm, extracting exposure time and gain values, injecting these parameters into a virtual load simulation system using a dynamic loading method, loading a preset ambient temperature gradient field and chip power consumption curve mapping, and obtaining a preliminary simulation running state including the thermal conduction delay of sensor nodes. This imaging parameter optimization method for visual imaging improves the long-term availability and imaging quality of visual imaging systems in real-time, multi-scenario tasks.

Description

Technical Field

[0001] This invention relates to the field of electronic vision imaging technology, and more specifically to an imaging parameter optimization method for vision imaging. Background Technology

[0002] In visual imaging systems, the selection of imaging parameters (such as exposure time and gain value) directly affects image quality, operational efficiency, and system stability. As imaging tasks become increasingly real-time and multi-scenario-based, high-performance imaging devices generally rely on optimization algorithms to automatically generate new parameter configurations to adapt to dynamic changes in complex environments. However, current technologies still predominantly employ a "algorithm output is directly sent to hardware for execution" model in parameter usage, lacking a systematic evaluation of the feasibility, security, and stability of parameters in actual hardware environments.

[0003] At the hardware level, imaging sensors, processing chips, and their peripheral circuits are all limited by physical characteristics, such as thermal conduction delay, power consumption response curves, and heat dissipation boundary conditions. When the parameter combination output by the optimization algorithm results in excessively long exposure times or excessively high gains, the internal heat accumulation of the device may rise rapidly and non-linearly. Existing systems typically lack the ability to simulate ambient temperature gradient fields, chip power consumption curves, or heat propagation paths at sensor nodes, making it impossible to predict potential temperature anomalies in advance. When such parameters are directly applied, they often lead to local temperatures approaching or exceeding protection thresholds, triggering overheat protection mechanisms and forcing the imaging system to interrupt its task. Furthermore, existing systems lack the ability to collect real-time logs of abnormal interruptions, making it impossible to determine trends based on temperature changes at sensor nodes or interruption response timing. They also lack mechanisms for curve fitting of temperature fluctuations, critical alarm zone determination, and environmental interference correction, making it difficult for the system to identify the correlation between exposure time, gain value, and heat accumulation in a timely manner. This also results in a lack of data support for parameter adjustments, preventing the formation of a stable and reliable optimization loop. Regarding parameter tuning, existing methods generally rely solely on the results of the current optimization algorithm, failing to incorporate multi-dimensional screening based on key indicators such as historical temperature data and overheat protection threshold distance. Furthermore, they lack joint verification of the "alternative parameter set" with real-time task requirements and hardware load sampling results. Therefore, when the algorithm continuously outputs large-amplitude parameter changes within a short period, the system is prone to sudden temperature overruns or unstable operation. These problems are particularly pronounced in real-world business scenarios. For example, in low-light real-time imaging tasks, the optimization algorithm might prioritize maximizing the combination of exposure and gain to enhance image brightness. However, without prior assessment of heat accumulation trends, this could lead to a rapid increase in chip temperature, ultimately triggering overheat protection and causing a momentary system shutdown. Such failures not only disrupt the task flow but can also result in irrecoverable loss of critical image data, severely impacting the reliability and continuous availability of the imaging system. Summary of the Invention

[0004] The purpose of this invention is to provide an imaging parameter optimization method for visual imaging, thereby solving the problems existing in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an imaging parameter optimization method for visual imaging, comprising: S1, receiving parameter configurations output by an optimization algorithm, extracting exposure time and gain values, injecting them into a virtual load simulation system using a dynamic parameter loading method, loading a preset ambient temperature gradient field and chip power consumption curve mapping, and obtaining a preliminary simulation running state including the thermal conduction delay of sensor nodes; S2, based on the preliminary simulation running state, setting fixed-duration segmented monitoring and high-temperature stage acquisition encryption rules, simultaneously enabling adaptive acquisition interval adjustment and acquisition frequency upper limit locking, and synchronously initiating hardware interrupt response in the virtual load simulation system. S3. For the simulation operation log records, extract the data packets that are triggered by abnormal interruption and collect them in real time. Combine the sensor node distribution information and timestamp alignment method to obtain the real-time temperature data stream at the frequency of the encrypted collection when the temperature is higher than the preset threshold, and judge the initial trend of temperature monitoring. S4. Based on the initial trend of temperature monitoring, the peak fluctuation is analyzed by polynomial curve fitting technology. If the peak fluctuation exceeds the preset threshold critical alarm range, the environmental interference correction mechanism based on the air cooling heat dissipation attenuation curve is triggered to obtain the corrected temperature change data.

[0006] Preferably, step S1 includes receiving parameter configurations output by a preset optimization algorithm, extracting exposure time and gain values ​​from them and injecting them into a virtual load simulation system, loading a preset ambient temperature gradient field to calculate the temperature distribution matrix, obtaining a temperature distribution matrix; for the temperature distribution matrix, mapping the power consumption distribution of sensor nodes using a chip power consumption curve, obtaining the heat source intensity value at the sensor node, and determining the heat conduction path; if the heat conduction path exceeds a preset threshold, injecting noise interference simulation to adjust the heat source distribution by superimposing random noise signals, obtaining the corrected heat source intensity value, and determining the heat conduction delay duration; dynamically adjusting the injected gain value according to the heat conduction delay duration, loading a real-time calibration mechanism to perform parameter compensation by comparing delay deviations, obtaining simulation environment data including delay compensation; and using the simulation environment data, fusing a preset preliminary simulation running state to obtain the final index of the heat conduction delay of the sensor node.

[0007] Preferably, step S2 includes: acquiring fixed-duration segmented monitoring data and high-temperature stage acquisition encryption rule parameters from the preliminary simulation operation state; using an adaptive acquisition interval adjustment mechanism to fuse the upper limit lock value of the acquisition frequency to obtain a preliminary monitoring sequence; injecting noise interference compensation service into the virtual load simulation system for the preliminary monitoring sequence, obtaining the compensated sequence distribution, and determining the hardware interrupt response timing simulation start point; loading the exposure time thermal accumulation coefficient calculation process through the compensated sequence distribution, the calculation process uses an integral method to accumulate and sum the heat changes within the time period in the sequence distribution, where the upper limit of the integral is the lower limit of the sequence duration and the starting point is the thermal accumulation distribution matrix; if the thermal accumulation distribution matrix exceeds a preset threshold, injecting multi-scene illumination mutations, obtaining the simulation environment data after mutation injection, and determining the operation log recording trigger conditions; according to the operation log recording trigger conditions, fusing delay compensation injection service, and calibrating and adjusting the time delay deviation to obtain the final simulation operation log.

[0008] Preferably, step S3 includes extracting instantaneous data packets generated by abnormal interruptions from simulation operation log records, obtaining the encrypted frequency sampling sequence carried by the instantaneous data packets and the corresponding trigger time lock information; aligning the instantaneous data packets with timestamps according to a pre-established sensor distribution map and node position mapping to obtain the correspondence between the spatial position locked by each sensor node at the trigger time and the encrypted frequency sampling sequence; for stages where the temperature is higher than a preset threshold, filtering the corresponding encrypted frequency sampling sequence using the trigger time lock information, and obtaining the real-time temperature data stream of each node by decryption; mapping the real-time temperature data stream to spatial positions according to the sensor distribution map to determine the correlation matrix between the temperature value of each node and the node position mapping; and drawing an initial trend line of the temperature data stream on the sensor distribution map based on the correlation matrix to determine the initial trend of temperature monitoring.

[0009] Preferably, step S4 includes: monitoring the initial trend of temperature; obtaining a peak fluctuation sequence using polynomial curve fitting technology; using trend data points as input and calculating polynomial coefficients using the least squares method to output the peak fluctuation sequence; analyzing the fluctuation amplitude of the peak fluctuation sequence by calculating the maximum absolute value of the difference between adjacent points in the sequence to determine the peak fluctuation value; if the peak fluctuation value exceeds a preset threshold alarm range, triggering an environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve; adjusting the heat transfer parameters of the air-cooled heat dissipation attenuation curve by fusing wind speed data, whereby the wind speed data is multiplied by an attenuation factor and then superimposed onto the curve equation; extracting humidity-related interference components from the environmental interference correction mechanism to obtain humidity interference factors and obtain interference adjustment coefficients, which are calculated by dividing the humidity interference factors by the standard thermal conductivity; and processing temperature change data using the interference adjustment coefficients by multiplying the temperature change data by the interference adjustment coefficients to obtain corrected temperature change data.

[0010] Preferably, the process further includes S5: Based on the corrected temperature change data, and combined with historical temperature data, compare its distance to the overheat protection threshold. If the distance is less than a preset critical alarm standard, activate the alternative parameter switching process to determine an adjusted parameter alternative set including a reduced exposure time heat accumulation coefficient and a gain value current amplification model. Specifically, this includes calculating the distance between the corrected temperature data and the overheat protection threshold based on historical temperature data. The distance is obtained by subtraction to get the difference between the current value and the threshold, resulting in a distance index. For the distance index, if the distance index is less than the preset critical alarm standard, activate the alternative parameter switching process to extract the reduced exposure time heat accumulation coefficient from the process. The reduced exposure time heat accumulation coefficient is adjusted by multiplying and fusing the time variable to determine the initial adjustment parameter. Using the initial adjustment parameter, fuse the gain value current amplification model. The gain value current amplification model generates an amplified output by scaling the current input proportionally, obtaining an extended set including humidity-affected parameters. Through the extended set, determine the compatibility between parameters in the set to obtain an adjusted parameter alternative set.

[0011] Preferably, the method further includes S6: extracting a set of adjusted parameters matching the current real-time task from a preset parameter library, and obtaining the final adapted imaging system configuration through multi-dimensional stability verification and hardware load sampling synchronous comparison. Specifically, this includes extracting a set of adjusted parameters matching the current real-time task from a preset parameter library, fusing external variables through environmental adaptation adjustment, wherein the environmental adaptation adjustment is categorized according to variable type and matched and fused one by one to obtain a preliminary screening set; using the preliminary screening set, performing multi-dimensional stability verification, wherein the multi-dimensional stability verification verifies the fluctuation range of each parameter in the set, and fusing the verification results with hardware load sampling data to obtain synchronous comparison indicators.

[0012] Preferably, step S6 further includes determining the relationship between the indicators and the threshold by synchronously comparing the indicators. If the indicators exceed the preset threshold, a dynamic resource allocation process is activated. The dynamic resource allocation process selects supplementary parameters from the backup library to determine the extended adaptation parameters. The extended adaptation parameters are obtained and combined with real-time task fusion to generate the final output. The fusion integrates task requirements through parameter superposition to obtain the imaging system configuration.

[0013] Preferably, the process also includes S7: using the final adapted imaging system configuration, continuing to perform fixed-duration segmented monitoring and real-time trigger acquisition for abnormal interruptions; if a hardware interruption response timing anomaly is detected again during continuous operation, then the corresponding sensor node thermal conduction delay and power consumption curve mapping data in the simulation log record is traced back to adjust parameters and determine overall operational stability. Specifically, this includes using the final adapted imaging system configuration, performing fixed-duration segmented monitoring, and acquiring hardware interruption response timing data during continuous operation through real-time trigger acquisition for abnormal interruptions; for the hardware interruption response timing data, when an anomaly is detected, tracing back the corresponding sensor node thermal conduction delay and power consumption curve mapping data in the simulation log record, wherein the simulation log record is extracted from a preset system backup to obtain basic indicators for parameter fine-tuning; and through the basic indicators for parameter fine-tuning, fusing image signal calibration variables, wherein the image signal calibration variables are extracted from a preset library and matched item by item with indicator values ​​to determine an extended stability verification set.

[0014] Preferably, step S7 further includes acquiring an extended stability verification set and performing multi-dimensional comparisons. The multi-dimensional comparisons verify the fluctuation range of each parameter within the set and fuse hardware load sampling data to determine the overall operational stability index. If the overall operational stability index exceeds a preset threshold, the backup sensor node adjustment is activated, and supplementary parameters are selected from the backup library to obtain the optimized imaging system configuration.

[0015] As can be seen from the above technical solution, the present invention has the following beneficial effects:

[0016] This imaging parameter optimization method for visual imaging accurately identifies the initial temperature trend and potential dangerous peak changes of sensor nodes through segmented monitoring, real-time acquisition of abnormal interruptions, simulation of sudden changes in illumination in multiple scenes, and encrypted acquisition of temperature data. It also combines polynomial fitting and environmental interference correction mechanisms to correct thermal fluctuations in real time, thereby obtaining more accurate temperature state judgment results. Furthermore, by comparing historical temperature data and calculating the distance to the overheat protection threshold, it achieves early warning of thermal risks and automatically activates the alternative parameter switching process when a risk is triggered, forming a set of safety parameters including a reduced exposure time thermal accumulation coefficient and a gain current amplification model. Subsequently, based on real-time task requirements, stability verification, and hardware load sampling comparison, the system selects the most suitable configuration from the parameter library, effectively avoiding overheating, interruptions, or system instability caused by unreasonable parameters output by the algorithm in a single instance. During continuous operation, this invention can also dynamically fine-tune parameters based on the abnormal interruption response sequence and the thermal conduction delay and power consumption mapping information in the simulation log, achieving adaptive stable control. In summary, this invention significantly improves the safety, continuity, and reliability of the imaging parameter optimization process, effectively preventing the system from triggering a protection shutdown due to parameter overruns or thermal accumulation runaway in complex scenarios, thereby enhancing the long-term availability and imaging quality of the visual imaging system in real-time, multi-scenario tasks. Attached Figure Description

[0017] Figure 1 This is a flowchart of the imaging parameter optimization method for visual imaging according to the present invention. Detailed Implementation

[0018] 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.

[0019] like Figure 1As shown, this invention provides a technical solution: an imaging parameter optimization method for visual imaging, including S1, receiving parameter configurations output by an optimization algorithm, extracting exposure time and gain values, injecting them into a virtual load simulation system using a dynamic parameter loading method, loading a preset ambient temperature gradient field and chip power consumption curve mapping, and obtaining a preliminary simulation running state including sensor node thermal conduction delay; S2, based on the preliminary simulation running state, setting fixed-duration segmented monitoring and high-temperature stage acquisition encryption rules, simultaneously enabling adaptive acquisition interval adjustment and acquisition frequency upper limit locking, synchronously starting hardware interrupt response timing simulation and exposure time thermal accumulation coefficient calculation in the virtual load simulation system, and determining simulation running log records under multi-scene illumination mutation injection; S3, for the simulation running log records, extracting data packets triggered by abnormal interruptions, combining sensor node distribution information and timestamp alignment, acquiring real-time temperature data streams at the encrypted acquisition frequency for stages with temperatures higher than a preset threshold, and determining the initial trend of temperature monitoring; S4, using the initial trend of temperature monitoring... The process involves several steps: S1) 1) Using polynomial curve fitting to analyze peak fluctuations. If the peak fluctuation exceeds a preset threshold alarm range, an environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve is triggered to obtain corrected temperature change data. S2) Based on the corrected temperature change data, historical temperature data is compared with the overheat protection threshold. If the distance is less than the preset threshold alarm standard, the alternative parameter switching process is activated to determine an adjusted parameter alternative set, including a reduced exposure time heat accumulation coefficient and a gain value current amplification model. S3) An adjusted parameter alternative set matching the current real-time task is extracted from the preset parameter library. Through multi-dimensional stability verification and hardware load sampling synchronous comparison, the final adapted imaging system configuration is obtained. S4) Using the final adapted imaging system configuration, fixed-duration segmented monitoring and real-time trigger acquisition for abnormal interruptions continue. If a hardware interruption response timing abnormality is detected again during continuous operation, the corresponding sensor node thermal conduction delay and power consumption curve mapping data in the simulation log are traced back to adjust the parameters and determine the overall operational stability.

[0020] This implementation method is based on the potential thermal accumulation effect of visual imaging systems under different exposure parameters and ambient temperatures. It constructs a virtual load simulation system to reproduce the thermal conduction behavior, power consumption response characteristics, and the impact of sudden changes in illumination on hardware interrupt timing of sensor nodes. The system first dynamically loads exposure time and gain, allowing these parameters to change the power consumption model and thermal conduction delay distribution of sensor nodes in real time during the simulation. Based on this, segmented monitoring and encrypted acquisition rules during high-temperature stages enable precise tracking of temperature changes, and an adaptive acquisition interval algorithm ensures the continuity and effectiveness of data sampling under high-temperature conditions. The addition of hardware interrupt response timing simulation identifies response lags caused by changes in exposure parameters, while the calculation of the thermal accumulation coefficient of exposure time reveals the temperature rise trend under sudden changes in illumination. Subsequently, the system extracts data from the abnormal interruption trigger segments in the simulation log and establishes a temperature data stream by combining sensor layout and timestamp correction methods. Analysis of temperature peak fluctuations using polynomial fitting identifies abnormal thermal behavior. When fluctuations exceed limits, an environmental interference correction mechanism automatically invokes the air-cooling heat dissipation attenuation curve to reduce the impact of external environmental fluctuations on temperature rise analysis, thereby obtaining a more realistic temperature change trajectory. Based on the corrected data, the distance between the current temperature and the overheat protection threshold is compared to determine whether the alternative parameter switching process needs to be triggered. If a switch is required, an alternative set containing the adjusted exposure heat accumulation coefficient and gain current model is retrieved from the parameter library, and a multi-dimensional stability check is applied to ensure that the adjusted imaging parameters meet the load requirements of the real-time task. The final adapted configuration continues to be used for system operation, and when an interruption response anomaly is detected again, the parameters are fine-tuned by reviewing the heat conduction and power consumption mapping data in the logs to achieve continuous closed-loop optimization.

[0021] S1 includes receiving parameter configurations output by a preset optimization algorithm, extracting exposure time and gain values ​​from them and injecting them into a virtual load simulation system, loading a preset ambient temperature gradient field to calculate the temperature distribution matrix, obtaining a temperature distribution matrix; for the temperature distribution matrix, mapping the power consumption distribution of sensor nodes using a chip power consumption curve, obtaining the heat source intensity value at the sensor node, and determining the heat conduction path; if the heat conduction path exceeds a preset threshold, injecting noise interference simulation to adjust the heat source distribution by superimposing random noise signals, obtaining the corrected heat source intensity value, and determining the heat conduction delay duration; dynamically adjusting the injected gain value according to the heat conduction delay duration, loading a real-time calibration mechanism to perform parameter compensation by comparing delay deviations, obtaining simulation environment data including delay compensation; and using the simulation environment data, fusing the preset preliminary simulation running state to obtain the final index of the heat conduction delay of the sensor node.

[0022] In this embodiment, firstly, during the design and calibration phase of the imaging system, historical data from a large number of representative imaging tasks are used to perform offline calculations on different combinations of exposure time and gain values ​​using a preset optimization algorithm. A comprehensive evaluation standard is set according to image sharpness, noise level, and temperature rise. All combinations are sorted, and several sets of exposure time and gain values ​​that meet image quality requirements and have low temperature rise are selected from the sorting results as parameter configurations and stored in a parameter configuration table as the parameter configurations output by the preset optimization algorithm. During actual simulation, the preset optimization algorithm reads a set of exposure time and gain values ​​corresponding to the current imaging task from the parameter configuration table and sets the exposure time and gain values ​​accordingly. The exposure time and gain values ​​are sequentially written into the input parameter area of ​​the virtual load simulation system to complete the injection of exposure time and gain values. After the imaging system structure design is completed, the virtual load simulation system divides the actual chip structure, dividing the chip surface and interior into several discrete regions with a fixed spatial step size. Each region is defined as a temperature calculation unit. In environmental experiments, the chip is run for a long time under different ambient temperatures, and temperature values ​​at various locations are collected using a temperature acquisition device. Based on the collection results, the temperature change relationship under different environmental conditions is organized according to spatial location to form an environmental temperature gradient field, which is pre-stored in the virtual load simulation system in the form of a numerical table. The virtual load simulation system updates the initial temperature of each temperature calculation unit at each simulation time step based on the temperature change records at corresponding locations in the ambient temperature gradient field. It also calculates the heat transfer between adjacent temperature calculation units step-by-step, considering the chip structure connections, until the simulation time covers the preset imaging task duration. At the end of each simulation time step, the temperature values ​​of all temperature calculation units are arranged according to their spatial location, forming a temperature distribution matrix for that time period. The system retains the temperature distribution matrix sequence for the entire simulation period after the simulation ends. After obtaining the temperature distribution matrix, during the chip development phase, the system further refines the chip's performance in various scenarios. The operating current and power consumption under different combinations of exposure time and gain values ​​are measured. The measured power consumption values ​​are organized according to the combination of exposure time and gain values ​​to form a chip power consumption curve, and the curve is stored in the virtual load simulation system in the form of a numerical table. At each time step in the simulation process, the virtual load simulation system looks up the corresponding power consumption data in the chip power consumption curve according to the currently used exposure time and gain values, and combines the physical location of the sensor nodes in the chip to take the power consumption value in the temperature calculation unit where each sensor node is located as the heat source intensity value of the temperature calculation unit, and generate the heat source intensity distribution at each sensor node in the simulation time step.Next, based on the chip structure and connection relationships, the virtual load simulation system starts from the sensor node with a larger heat source intensity value and searches for adjacent temperature calculation units step by step along the actual heat conduction path of the chip material, according to the physical distance and the material's heat conduction direction. Continuous connections from nodes with larger heat source intensity values ​​to nodes with smaller heat source intensity values ​​are marked as heat conduction paths. The system then calculates the number of sensor nodes on each heat conduction path, the spatial length traversed by the path, and the cumulative time difference of temperature change along the path direction. These statistical results are compared with a threshold predetermined during the chip reliability design phase. This threshold is determined through multiple rounds of long-term high-temperature operation tests during the chip reliability design phase, specifically by recording different heat conduction path lengths under continuous operation conditions. The chip failure probability corresponding to the same number of nodes and different temperature change time differences is determined based on the chip's maximum allowable failure rate. The most stringent limit among the three indicators—heat conduction path length, number of nodes, and temperature change time difference—is selected before the failure rate reaches the upper limit. This limit is set as a preset threshold and fixed in the simulation system. When the statistical result of any heat conduction path exceeds the preset threshold, the virtual load simulation system initiates a noise interference simulation process on the temperature calculation unit involved in that path. During the chip design phase, based on factors such as power fluctuations, process discreteness, and environmental disturbances in the actual operating environment, a large amount of measured temperature data is statistically analyzed to obtain the fluctuation range and frequency of the heat source intensity. The frequency, representing the intensity range and occurrence frequency of random noise signals, is used in the noise interference simulation process. The virtual load simulation system generates a random noise value within a predetermined intensity range for each temperature calculation unit along the path at each simulation time step. This random noise value is then superimposed on the heat source intensity value of the corresponding temperature calculation unit to generate a corrected heat source intensity value. The temperature distribution is then recalculated using this corrected heat source intensity value to obtain the temperature distribution matrix corrected for noise interference. Subsequently, the virtual load simulation system performs time analysis on the corrected temperature distribution matrix sequence. For each heat conduction path, starting from the time point when the heat source intensity value begins to increase significantly, the system searches progressively forward along the simulation time until the end node of the path. Once the temperature reaches a stable state, the time difference between the point when the heat source begins to act and the point when the temperature at the end node of the path reaches a stable state is defined as the heat conduction delay duration. The heat conduction delay duration corresponding to each heat conduction path is recorded. After obtaining the heat conduction delay duration, the virtual load simulation system adjusts the gain value injection process based on the deviation between the heat conduction delay duration of each heat conduction path and the target delay duration preset according to the target response speed during the imaging system design phase. During the setting of the target delay duration, multiple sets of different gain value combinations are tested on the chip under laboratory conditions, and the time required for the image signal to reach stable output is recorded. The shortest time to meet the imaging quality is fixed in the system as the target delay duration.During the simulation, when the actual heat conduction delay of a path exceeds the target delay, the virtual load simulation system adjusts the gain value of the corresponding path's region downwards by a pre-set decreasing step size according to the magnitude of the deviation. The adjusted gain value is then reinjected in the next simulation round, and this simulation and comparison process is repeated until the deviation between the actual heat conduction delay and the target delay is reduced to a preset allowable range. In each comparison, the virtual load simulation system uses the difference between the actual heat conduction delay and the target delay as the delay deviation. A real-time calibration mechanism controls the direction and step size of the gain value adjustment based on the delay deviation. Through multiple iterations, simulation environment data containing the gain value adjustment results and the corresponding temperature distribution results is generated. After obtaining the simulation environment data, the virtual load simulation system fuses the simulation environment data, including delay compensation, with the generated and stored preliminary simulation running state. It compares the heat conduction delay time of each sensor node in the preliminary simulation running state with the corresponding heat conduction delay time in the simulation environment data including delay compensation, calculates the difference between the two, and combines this difference with the value in the preliminary simulation running state to obtain a comprehensive index that reflects both the original heat conduction characteristics and the compensation effect. This comprehensive index includes the heat conduction delay time of each sensor node and the average heat conduction delay time of the entire system. The virtual load simulation system outputs this comprehensive index as the final indicator of the sensor node's heat conduction delay.

[0023] S2 includes obtaining fixed-duration segmented monitoring data and high-temperature stage acquisition encryption rule parameters from the preliminary simulation operation status, and using an adaptive acquisition interval adjustment mechanism to fuse the upper limit lock value of the acquisition frequency to obtain a preliminary monitoring sequence; for the preliminary monitoring sequence, injecting noise interference compensation service into the virtual load simulation system to obtain the compensated sequence distribution and determine the hardware interrupt response timing simulation start point; through the compensated sequence distribution, loading the exposure time thermal accumulation coefficient calculation process, the calculation process uses an integral method to accumulate and sum the heat changes within the time period in the sequence distribution, where the upper limit of the integral is the lower limit of the sequence duration as the starting point, to obtain the thermal accumulation distribution matrix; if the thermal accumulation distribution matrix exceeds a preset threshold, injecting multi-scene illumination mutations, obtaining the simulation environment data after mutation injection, and determining the operation log recording trigger conditions; based on the operation log recording trigger conditions, fusing the delay compensation injection service, and calibrating and adjusting the time delay deviation to obtain the final simulation operation log.

[0024] In this embodiment, firstly, based on the obtained preliminary simulation running state, which is a time-series data set formed by recording temperature, power consumption, and interruption status at fixed intervals within a preset simulation duration, the fixed duration of each monitoring segment is predetermined during the configuration phase according to the total duration of the imaging task and the monitoring granularity allowed by the imaging system. For example, the total duration is evenly divided into several consecutive time periods, and the fixed duration of each monitoring segment is given in the configuration file in the form of specific values, which are directly called at the start of the simulation. Thus, the temperature and interruption records corresponding to each segment are segmented and extracted from the preliminary simulation running state in chronological order. This generates segmented monitoring data with fixed durations. The high-temperature phase data acquisition encryption rules are established through long-term operation of the imaging system under various ambient temperature conditions and imaging loads during chip prototype environmental testing. The highest temperature value, temperature change rate, and image quality changes of each sensor node during continuous operation are statistically analyzed. The temperature value before a significant increase in image noise or approaching overheat protection is set as the starting threshold for the high-temperature phase. The ratio of the number of samples required to accurately reflect temperature changes during the high-temperature phase to the number of samples during the normal phase is set as the sampling encryption factor. These two specific values ​​and their corresponding applicable temperature ranges constitute the high-temperature phase data. The temperature range is monitored during the high-temperature phase. The average temperature of the current time period is compared directly with the starting threshold of the high-temperature phase during simulation. When the temperature reaches or exceeds this threshold, the number of samples within that time period is increased by the sampling encryption factor, resulting in denser sampling points than in the normal phase. The adaptive sampling interval adjustment mechanism uses the absolute difference in temperature change between two consecutive time periods and the number of interruption events as the criteria. If the absolute difference in temperature change is greater than the preset temperature change threshold or the number of interruption events is greater than the preset interruption density threshold, the sampling interval is shortened by a preset ratio in the next monitoring time period; otherwise, it is appropriately extended by a preset ratio. The temperature change threshold and interruption density threshold are also obtained through long-term statistical analysis of the prototype and are fixed in the simulation configuration in specific numerical form. To avoid excessive data processing pressure due to overly frequent sampling, the upper limit of the sampling frequency is determined during the configuration phase based on hardware processing capabilities and data bandwidth test results. This value represents the maximum number of samples allowed per unit time. When the sampling frequency calculated after adaptive adjustment exceeds this value, it is automatically truncated to the upper limit of the sampling frequency. Thus, the final determined sampling interval and number of samples are obtained for each monitoring time period on the time axis. These are arranged in chronological order to obtain the preliminary monitoring sequence.To eliminate the impact of environmental noise and simulation errors on temperature and interruption data in the initial monitoring sequence, during the noise interference compensation configuration phase, a real imaging system was first run under stable environmental conditions to record temperature change curves and interruption recording curves when there was no significant external interference. These curves were used as the baseline sequence. Then, the system was run under actual operating conditions with wind interference and power fluctuations, recording monitoring sequences under the same task. The two types of sequences were subtracted point-by-point at the same time position to obtain the typical offset range caused by environmental noise and the probability distribution of offset occurrence. The offset ranges and their occurrence probabilities under different temperature ranges and different interruption levels were compiled into a noise characteristic parameter table. The noise characteristic parameters include the lower limit, upper limit, and average compensation offset for each temperature range, as well as the compensation application ratio. During simulation, for each sampling point in the initial monitoring sequence, the corresponding temperature range was found based on the temperature value of that sampling point. The average compensation offset for that range was selected from the noise characteristic parameter table, and the compensation application ratio was calculated. For example, the current sampled value is corrected. If the temperature is too high, the average compensation offset is subtracted; if the temperature is too low, the average compensation offset is added. The interrupt event count is corrected in the same way. After processing all sampled points in sequence, the compensated sequence distribution is obtained. In the compensated sequence distribution, in order to determine the starting point of the hardware interrupt response timing simulation, the interrupt response start temperature threshold and the interrupt surge threshold are set in advance during the configuration stage. The interrupt response start temperature threshold is determined by recording the temperature value when the interrupt response starts to lag significantly during the test. The interrupt surge threshold is determined by statistically analyzing the maximum number of normal interrupts in each time period and multiplying it by a safety factor. During the simulation, the compensated sequence distribution is scanned sequentially from the starting point of time. Once the average temperature of a certain time period exceeds the interrupt response start temperature threshold and the increase in the number of interrupts in that time period exceeds the interrupt surge threshold, the starting time of that time period is marked as the hardware interrupt response timing simulation starting point. Subsequent timing simulations will then perform fine simulation and recording of the trigger time and response time of each interrupt signal from that moment.In the calculation of the exposure time thermal accumulation coefficient, this coefficient was determined through extensive experiments during the chip thermal calibration phase. Specifically, while keeping other conditions constant in the experimental environment, only the exposure time configuration was changed, and the temperature rise per unit time of the sensor node was recorded for each exposure time. A one-to-one correspondence table was formed between the temperature rise per unit time and the specific exposure time value, which is the exposure time thermal accumulation coefficient. During simulation, for each monitoring time period in the compensated sequence distribution, the exposure time value used in that time period is first read, and then the corresponding heat increase per unit time is found according to the exposure time thermal accumulation coefficient table. The heat accumulation value for a given time period is calculated by multiplying the increase in heat per unit time by the fixed duration of that time period. Then, starting from the initial monitoring period, the heat accumulation value for each segment is added sequentially until the entire initial monitoring sequence ends. This segment-by-segment accumulation method performs an integral calculation of the heat change within the sequence duration. The accumulated heat value at the end of each time period is correlated with the sensor's spatial location to form a heat accumulation distribution matrix. The preset threshold is determined during the imaging system design and reliability verification phase based on the chip's maximum allowable heat accumulation level. Specifically, it involves continuously increasing the imaging load and exposure time combination under constant environmental conditions, causing the values ​​in the heat accumulation distribution matrix to gradually increase. Simultaneously observe the image noise, number of bad pixels, and hardware overheat protection triggering status of the chip output. When any indicator reaches the system's allowed upper limit, record the heat accumulation value at each position in the heat accumulation distribution matrix. The smallest of these values ​​is taken as the global safety upper limit heat accumulation value and set as a preset threshold. This ensures that if any position in the heat accumulation distribution matrix exceeds this value, there is a potential overheating risk. During simulation, after each round of heat accumulation distribution calculation, the system compares the heat accumulation values ​​at all positions in the matrix with the preset threshold. Once any value at any position exceeds the preset threshold, a multi-scene illumination mutation injection process is triggered. During implementation, a lighting scene library is pre-established, which consists of actual collected indoor scenes, outdoor strong light scenes, shadow scenes, and high dynamic range scenes. Each scene contains corresponding light intensity curves and contrast information. During the simulation configuration phase, a scene mutation triggering order and switching time offset range are configured for each scene. When the thermal accumulation distribution matrix exceeds a preset threshold, the system selects the next lighting scene from the scene library in a pre-set order, switches the light intensity and contrast of the current simulation to the light intensity and contrast of the new scene within a set very short time, and updates the corresponding exposure time value and gain value input, thereby obtaining the simulation environment data after mutation injection.The trigger conditions for running the log are determined during the configuration phase based on fault analysis requirements. These conditions fall into three categories: first, the number of times the heat accumulation distribution matrix exceeds a preset threshold within a given time period reaches a preset threshold; second, the interrupt response time monotonically increases over several consecutive monitoring time periods; and third, after a sudden change in lighting conditions, the absolute value of the temperature change exceeds a preset temperature change threshold within a short time window. These threshold values, time window lengths, and temperature change thresholds are all statistically derived from historical experimental data and then stored in the system. During simulation, after each round of calculation, the system compares the current result with each of the above conditions. When any condition is met, running the log is started, writing the temperature value, heat accumulation value, lighting scene identifier, exposure time value, gain value, and interrupt response time for each monitoring time period after the trigger time into the running log in chronological order. To ensure that the time information recorded in the running log accurately reflects the sensor and interrupt response... Based on the actual latency, latency compensation injection is integrated when generating the final simulation run log. The time delay deviation used in the latency compensation injection is a value obtained by comparing the heat conduction delay duration with the target delay duration. When recording each event in the simulation, the corresponding time delay deviation is added to the original simulation time of that event as a calibrated timestamp. After each calibration, the time delay deviation is checked again to see if it is within the preset allowable range. If it is still too large, the gain value or other parameters related to heat conduction are adjusted according to the deviation in the next round of simulation, and the time delay deviation is recalculated. This process is iterated until the time delay deviation is less than the allowable value. All event records using calibrated timestamps are combined into the final simulation run log. Each record in the final simulation run log contains temperature information, heat accumulation information, and interruption response information after noise compensation, multi-scene illumination mutation injection, and latency compensation injection.

[0025] S3 includes extracting instantaneous data packets triggered by abnormal interruptions from simulation operation logs, obtaining the encrypted frequency sampling sequence and corresponding trigger time lock information carried by the instantaneous data packets; aligning the instantaneous data packets with timestamps based on a pre-established sensor distribution map and node location mapping to obtain the correspondence between the spatial location of each sensor node locked at the trigger time and the encrypted frequency sampling sequence; for stages where the temperature is higher than a preset threshold, filtering the corresponding encrypted frequency sampling sequence using the trigger time lock information, and obtaining the real-time temperature data stream of each node through decryption; mapping the real-time temperature data stream to spatial locations according to the sensor distribution map to determine the correlation matrix between the temperature value of each node and the node location mapping; and drawing an initial trend line of the temperature data stream on the sensor distribution map based on the correlation matrix to determine the initial trend of temperature monitoring.

[0026] In this embodiment, the instantaneous data packets triggered by abnormal interruptions are first extracted from the simulation operation log records. The simulation operation log records are a collection of records saved in chronological order in the aforementioned steps. Each record includes at least an interruption type flag, the event occurrence time, the sampled value of the corresponding sensor node, and an encrypted frequency sampling sequence identifier generated according to encryption rules when a high-temperature phase occurs. An abnormal interruption refers to an interruption event in the system calibration phase that, through multiple rounds of simulation experiments and actual machine operation experiments, has a response time exceeding the upper limit of the normal response time or an interruption count exceeding the upper limit of the normal interruption count per unit time. The upper limit of the normal response time is obtained by adding a fixed safety margin to the average of multiple measured response times, and the upper limit of the normal interruption count is obtained by multiplying the maximum value of multiple measured interruption counts per unit time by a safety factor. An interruption event that simultaneously meets any of the above conditions is considered an abnormal interruption. The system marks each interrupt as an abnormal interrupt. During runtime, the system scans all interrupt records in the simulation run log in chronological order. For each record marked as an abnormal interrupt, it reads its associated instantaneous data packet. The instantaneous data packet consists of multiple temperature and interrupt sampling data collected at high frequency within a fixed time window before and after the abnormal interrupt is triggered. This time window is set with specific start and end offsets during the configuration phase, based on the simulation time, for example, within a range of several microseconds before and after the trigger time. In addition to the sampled values, the instantaneous data packet also contains a fixed encrypted frequency sampling sequence and trigger time lock information used to determine the time base. The system reads each instantaneous data packet sequentially, parses the encrypted frequency sampling sequence into a sequence of sampled values ​​arranged at fixed time intervals, and uses the trigger time lock information as the time zero point of all sampling records in the data packet.Then, the instantaneous data packets are timestamped according to the pre-established sensor distribution map and node position mapping. The sensor distribution map is a planar or spatial coordinate map drawn according to the actual wiring and pixel array physical layout after the imaging sensor design is completed. Each sensor node has a definite coordinate position and unique number in the distribution map. The node position mapping is a mapping table that maps the node number used during simulation to the corresponding coordinate position in the sensor distribution map. During simulation and recording, each sampling record carries the node number and the time offset relative to the trigger time. The specific calculation process of timestamp alignment is as follows: For each sampling record in each instantaneous data packet, the time offset relative to the trigger time in the record is added to the absolute time value in the trigger time lock information in the data packet to obtain the absolute timestamp of the sampling record on the entire simulation time axis. Then, the corresponding sensor node spatial coordinates are found in the node position mapping table according to the node number in the sampling record. This coordinate is associated with a certain sampling value in the encrypted frequency sampling sequence corresponding to the node at that time. After the above traversal processing, the correspondence between the spatial position of each sensor node and the encrypted frequency sampling sequence in the vicinity of the trigger time is obtained.For stages where the temperature exceeds a preset threshold, the preset threshold is a temperature value determined from measured data during the system's thermal safety calibration phase. During calibration, the imaging system is run for extended periods under varying ambient temperatures and loads, recording changes in image noise, the number of bad pixels, and the number of abnormal interruptions as the temperature gradually increases. When a certain temperature level begins to significantly degrade image quality or significantly increase the number of abnormal interruptions, this temperature level is recorded as a risk temperature. The highest temperature slightly below the risk temperature that still does not show obvious anomalies is selected from multiple test samples as the preset threshold. This threshold is stored as a specific numerical value in the simulation system. During runtime, the system calculates the temperature statistics within a time window near the trigger time for each instantaneous data packet. A representative temperature for this window can be obtained by averaging all temperature samples within the window. This representative temperature is compared with the preset threshold. When the representative temperature exceeds the preset threshold, the time window is marked as a stage where the temperature exceeds the preset threshold. The system then uses trigger time locking information to filter encrypted frequency sampling sequences falling within this stage from the instantaneous data packets, only... The encrypted frequency sampling sequences corresponding to these stages are retained for subsequent processing. The encrypted frequency sampling sequences are temperature data collected at a sampling frequency higher than that of ordinary monitoring during the high-temperature stage. To prevent tampering or leakage during transmission and storage, the values ​​are transformed and rearranged according to pre-defined encryption rules during recording. These encryption rules are fixed in the system design and include segmenting the original temperature sampling values ​​into segments of fixed length, inserting redundant markers for verification in each segment, and adding or subtracting a fixed offset or scaling up or down by a fixed ratio for each sampling value. The decryption process is the reverse calculation process of the encryption rules. The system first identifies the start and end positions of each segment based on the redundant markers, removes the redundant markers and restores the original segmented structure, and then performs a reverse transformation on each value according to the recorded offset and ratio, restoring the offset and scaling up the ratio to restore the original temperature value. Finally, a real-time temperature data stream of each node is obtained according to the high-frequency sampling time. Each real-time temperature data stream contains the temperature value of the node at several consecutive sampling times and the time offset of these times relative to the trigger time.Next, the real-time temperature data stream is spatially mapped according to the sensor distribution map. The calculation process for spatial location mapping is as follows: For each node's real-time temperature data stream, the spatial coordinates of that node are read from the node location mapping table. A correspondence is established between the node's temperature value at each sampling time and the fixed spatial coordinates. The temperatures and spatial coordinates of all nodes at the same sampling time are combined into a set of spatial temperature distribution data. By arranging these spatial temperature distribution data in sequence along the time dimension, an association matrix is ​​constructed to map the temperature values ​​of each node to the node locations. The association matrix can be understood as a temperature data table with dual indexes of time and nodes. In this data table, each row corresponds to a sampling time, each column corresponds to a sensor node, and the elements in the table are the temperature values ​​of that node at that time. Finally, an initial trend line of the temperature data stream on the sensor distribution map is drawn based on the association matrix. The process of drawing the initial trend line is as follows: At each sampling time, the nodes with the largest temperature values ​​are found from the association matrix, and the spatial coordinates of these nodes on the sensor distribution map are marked as high. The above operation is repeated at the next sampling time. The positions of the corresponding high-temperature points at two adjacent sampling times are then connected by straight lines on the spatial distribution map. By continuously connecting these high-temperature point positions throughout the entire analysis period, a curve extending continuously on the sensor distribution map is formed. This curve reflects the direction of movement and expansion range of the high-temperature area on the sensor plane. At the same time, the system can also calculate the temperature change rate based on the change of the average temperature value of all nodes at each sampling time in the correlation matrix over time. The temperature change rate sequence is obtained by dividing the difference of the average temperature value of multiple adjacent time periods by the corresponding time interval. When the temperature change rate remains positive for multiple consecutive time periods and the value is greater than the temperature rise rate threshold obtained in advance based on experimental statistics, the initial trend line is determined to be a continuous upward trend, indicating that the initial trend of temperature monitoring is an expansion of temperature. When the temperature change rate changes from a positive value to close to zero or a negative value in a short period of time and the range of the high-temperature area on the sensor distribution map of the initial trend line does not continue to expand, it is determined that the initial trend of temperature monitoring is stabilizing or declining.

[0027] S4 includes monitoring the initial temperature trend, obtaining a peak fluctuation sequence using polynomial curve fitting technology. This technology uses trend data points as input, calculates polynomial coefficients using the least squares method, and outputs the peak fluctuation sequence. For the peak fluctuation sequence, the fluctuation amplitude is analyzed by calculating the maximum absolute value of the difference between adjacent points in the sequence to determine the peak fluctuation value. If the peak fluctuation value exceeds a preset threshold alarm range, an environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve is triggered. The air-cooled heat dissipation attenuation curve adjusts the heat transfer parameters by fusing wind speed data. This fusing involves multiplying the wind speed data by an attenuation factor and then superimposing it onto the curve equation. According to the environmental interference correction mechanism, humidity-related interference components are extracted from the mechanism to obtain the humidity interference factor and an interference adjustment coefficient. This interference adjustment coefficient is calculated by dividing the humidity interference factor by the standard thermal conductivity. The temperature change data is processed using the interference adjustment coefficient by multiplying the temperature change data by the interference adjustment coefficient to obtain the corrected temperature change data.

[0028] In this embodiment, the system first represents the initial temperature monitoring trend obtained from the previous steps as a set of trend data points arranged in chronological order. Each trend data point consists of a specific time value and a specific temperature value. During the system design and calibration phase, the developers statistically analyzed the fitting errors of various polynomial fitting forms from first to fourth order based on a large amount of historical operating data. They compared the magnitude of the fitting error and whether excessive fluctuations occurred under different orders. The third-order polynomial, which had a smaller fitting error and whose fitting curve shape could stably reflect the temperature change trend, was determined as the fixed fitting order. The third order was written as a fixed parameter into the configuration file and would not be adjusted during runtime. When performing polynomial curve fitting, the system uses all trend data points as input to construct a set of polynomial coefficients. Initially, all coefficient values ​​are set to zero. Then, a fitting calculation is performed for each data point in the trend dataset. Specifically, the fitted temperature corresponding to the current coefficient value is calculated, and the actual temperature of the data point is subtracted from the fitted temperature to obtain a difference. The square of this difference is accumulated into the total error. After traversing all trend data points, the total error value of this round of fitting is obtained. During the calibration phase, a convergence error threshold is pre-determined based on historical data. This error threshold is the maximum allowable error when image quality is not affected and the temperature trend characteristics are correctly reflected. Allowing for error, the error threshold is written into the system configuration; during the fitting process, the system adjusts the polynomial coefficients using the principle of minimum error. Specifically, it calculates the direction and step size for increasing or decreasing each coefficient based on the sensitivity of the total error to changes in each coefficient. Before the next round of traversal, all coefficients are updated simultaneously, and the total error is recalculated by traversing all trend data points again. This process is repeated iteratively. When the total error is first less than or equal to the pre-set convergence error threshold, or when the change in the total error is lower than the minimum error change judgment value set in the calibration stage in multiple consecutive iterations, the system stops iterating and fixes the coefficients of each polynomial obtained at this time as the final fitting coefficients. After fitting, the system moves point by point from the start time to the end time within the time range covered by the initial temperature monitoring trend, with a fixed time step. At each time step, the system calculates the fitting temperature value corresponding to the polynomial using the final fitting coefficients, forming a continuous and smooth fitting temperature curve. Then, the system searches for local peak points in the fitting temperature curve by sequential scanning. Specifically, the fitting temperatures of any three adjacent time steps are compared. If the fitting temperature of the middle time step is greater than the fitting temperatures of the previous and next time steps, the time value and fitting temperature value of the middle time step are combined and recorded as a peak point. The system records all peak points in chronological order, forming a peak fluctuation sequence. Each element in this sequence consists of a time value and a peak temperature value.Subsequently, the system analyzes the fluctuation amplitude of the peak fluctuation sequence. Starting from the first peak point in the sequence, the system reads the current peak temperature and the next peak temperature in sequence. It performs a difference calculation on these two temperature values, subtracts the smaller temperature value from the larger temperature value to obtain the temperature difference between the two peaks, and then uses the absolute value of the temperature difference as the fluctuation amplitude of this pair of adjacent peaks. The system sequentially traverses the entire peak fluctuation sequence, calculates the fluctuation amplitude between all adjacent peaks in turn, and stores each fluctuation amplitude value in a fluctuation amplitude list. After all calculations are completed, the system finds the fluctuation amplitude with the largest value in the list and defines the peak fluctuation value obtained in this analysis. The preset threshold alarm range is determined during the system thermal stability calibration and reliability testing phase. Specifically, multiple different environmental conditions are set up in the laboratory, including different ambient temperatures, wind speeds, and humidity levels. The imaging system is driven to run for a long time with multiple representative imaging tasks. Under each condition, the initial temperature monitoring trend is continuously recorded, and the peak fluctuation sequence and corresponding peak fluctuation value are calculated according to the above method. At the same time, the image quality detection program and hardware fault monitoring program record whether there are faults such as severe image noise, obvious stripes, overheat protection triggering, or frequent abnormal interruptions under each condition. After all conditions are statistically analyzed, the minimum and maximum values ​​of the peak fluctuation values ​​corresponding to all conditions without the above faults are found. The minimum value is determined as the lower limit of the threshold alarm range, and the maximum value is determined as the upper limit of the threshold alarm range. These two specific values ​​are fixedly stored in the configuration file as the standard for determining whether to trigger an alarm. In actual operation, when the calculated peak fluctuation value is less than the lower limit, it indicates that the temperature fluctuation is too small but within a safe range. When the peak fluctuation value is greater than the upper limit, it indicates that the temperature fluctuation exceeds the maximum empirical value of the system under safe operating conditions. The system immediately judges this as an abnormal fluctuation and triggers an environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve. The air-cooled heat dissipation attenuation curve is obtained through a dedicated air-cooling experiment during the system thermal calibration phase. During calibration, the imaging system is placed in an air duct with adjustable wind speed, and several fixed wind speed levels are set. Each wind speed level corresponds to a specific wind speed value. At each wind speed level, the system is started with a uniform initial temperature and uniform workload. The temperature values ​​of key chip locations are recorded at fixed time intervals. The temperature drop in each time interval is calculated, and the temperature drop is divided by the corresponding time interval length to obtain the actual heat conduction capacity at that wind speed and time period. The heat conduction capacity values ​​of multiple time periods and multiple wind speed levels are organized into scatter plot data corresponding to time and heat conduction capacity. Then, an air-cooled heat dissipation attenuation curve table changing with time is generated using interpolation or fitting methods. This table can be directly consulted in subsequent operation.During the calibration process described above, an attenuation factor is also calculated. This attenuation factor is a proportional constant describing the intensity of the influence of wind speed changes on heat transfer capacity. It is determined by calculating the change in heat transfer capacity corresponding to a unit change in wind speed at all wind speed levels, averaging the changes of all samples to obtain a stable ratio, and fixing this ratio as the attenuation factor. During operation, when the environmental interference correction mechanism is triggered, the system first reads the wind speed data recorded in the current environmental configuration. This wind speed data is provided by on-site testing or a fixed configuration during system deployment and is a fixed value. Then, at each temperature change time step, the system looks up the basic heat transfer parameter corresponding to that time step from the air-cooled heat dissipation attenuation curve table. By multiplying the current wind speed value by the attenuation factor, a wind speed correction is obtained. This wind speed correction is then directly added to the basic heat transfer parameter to form a corrected heat transfer parameter that includes the influence of wind speed. This process is equivalent to multiplying the wind speed data by the attenuation factor and then superimposing it onto the air-cooled heat dissipation attenuation curve. Besides wind speed, ambient humidity also interferes with heat transfer capacity. Therefore, the system pre-stores humidity-related interference data in its environmental interference correction mechanism. This interference data is obtained during the calibration phase by conducting multiple sets of experiments while changing the relative humidity of the environment and keeping other conditions constant. Specifically, it records the heat transferred per unit time under the same temperature difference and the same geometric structure under different humidity conditions. These experimental results are compared with the heat transfer results under standard humidity conditions to obtain the amount of heat transfer enhancement or reduction under each humidity condition. These enhancements or reductions are divided by the heat transfer results under standard humidity conditions to obtain a set of humidity interference factors. These factors are stored in a humidity interference data table according to humidity values. During operation, the system looks up the corresponding humidity interference factor in the humidity interference data table based on the current ambient humidity value and uses this humidity interference factor as the humidity interference factor. The standard thermal conductivity is a baseline thermal conductivity obtained through repeated experiments in a dry standard environment. During calibration, the ambient humidity is kept below a preset upper limit, a fixed temperature difference and a fixed heat transfer path size are set, and the heat transferred through the path per unit time is measured. This process is repeated multiple times, and the average value is calculated. This average value is used as the standard thermal conductivity and is fixed in the system. The calculation process of the interference adjustment coefficient is as follows: divide the value of the humidity interference factor found under the current environmental conditions by the standard thermal conductivity value to obtain a dimensionless proportional value, which is the interference adjustment coefficient. This coefficient directly reflects the change ratio of the thermal conductivity under the current humidity conditions relative to the standard conditions. When the coefficient is greater than 1, it means that the actual thermal conductivity is higher than the standard conditions. When the coefficient is less than 1, it means that the actual thermal conductivity is lower than the standard conditions.Finally, the system uses an interference adjustment coefficient to process the temperature change data. For each temperature change data point that needs correction, the original temperature change is read point by point in the time series. The original temperature change is multiplied by the interference adjustment coefficient to obtain the corrected temperature change considering humidity interference. If the heat conduction parameters have already been corrected using the wind-cooled heat dissipation attenuation curve, the temperature change data itself already reflects the wind speed effect. At this time, the interference adjustment coefficient further compensates for the humidity effect. The system repeats the above multiplication calculation for all time steps within the entire monitoring period to obtain a corrected temperature change data sequence covering the entire monitoring period. This sequence is used as the sole temperature input for subsequent steps to determine the distance to the overheat protection threshold and to execute parameter switching strategies, ensuring that the temperature assessment and alarm determination results are reliable and repeatable even in the presence of wind speed and humidity interference.

[0029] S5 includes calculating the distance between the calibrated temperature data and the overheat protection threshold based on the historical temperature data. The distance is obtained by subtraction to get the difference between the current value and the threshold, resulting in a distance index. If the distance index is less than a preset critical alarm standard, the alternative parameter switching process is activated. The process extracts a heat accumulation coefficient for reducing exposure time. This coefficient is adjusted by multiplying and fusing time variables to determine preliminary adjustment parameters. Using the preliminary adjustment parameters, a gain value current amplification model is fused. The gain value current amplification model generates an amplified output by scaling the current input, obtaining an extended set containing humidity-affected parameters. The compatibility between parameters in the extended set is determined to obtain a candidate set of adjusted parameters.

[0030] In this embodiment, the calibration temperature data obtained in the previous stage and the historical temperature data corresponding to the current imaging task type are read first. The calibration temperature data is the actual operating temperature sequence arranged in chronological order after environmental interference correction for each sensor node under the same task. The historical temperature data is the stable operating temperature sequence recorded and compiled in long-term reliability tests without overheating protection or severe image degradation. The overheating protection threshold is determined in the system design stage through high-load continuous operation tests. In the test, the imaging load is gradually increased and the continuous operation time is extended. The temperature change of the core area of ​​the chip is monitored in real time. When the hardware overheating protection circuit is detected to be triggered and the test results are consistent in a short period of time, the lower limit of the highest operating temperature corresponding to the hardware protection action at this time is fixed as the overheating protection threshold and written into the system configuration file with a single specific value. During operation, the system iterates through the calibration temperature data at fixed time steps. At each time step, it performs a subtraction operation between the current calibration temperature value and the overheat protection threshold. The temperature margin at that time point is obtained by subtracting the current temperature value from the overheat protection threshold value. Under the same task type, the system also extracts the temperature value corresponding to the current time position from the historical temperature data. By comparing the difference between the two, it determines whether the current operation is in the period of the fastest temperature rise in history. The minimum value of the temperature margin among all time points in the period of the fastest temperature rise is determined as the distance index for this round of evaluation. This distance index represents the most unfavorable safety margin of the current operation relative to the overheat protection threshold in a single numerical form. The preset critical alarm standard is obtained through statistical analysis of a large number of operational samples during the system's thermal safety calibration phase. During the calibration process, multiple different operating conditions and task combinations are run for extended periods. In each operating condition, the distance index is calculated using the method described above, and it is recorded whether significant image degradation or a boundary state before hardware protection occurs under that condition. Then, the corresponding distance index is extracted from all operating conditions that are in a boundary state but have not actually triggered protection. The largest value among these indices is fixed as the critical alarm standard, ensuring that this standard accurately represents the critical margin for transitioning from safe operation to overheating risk. During real-time operation, the system compares the current distance index with the critical alarm standard. When the distance index is greater than or equal to the critical alarm standard, the existing parameters remain unchanged; when the distance index is less than the critical alarm standard, the system immediately enters the alternative parameter switching process.In the alternative parameter switching process, the system first retrieves the original exposure time thermal accumulation coefficient corresponding to the current exposure time from the parameter management. This coefficient is obtained during the thermal calibration phase by measuring and normalizing the temperature rise caused by exposure per unit time under different exposure time configurations, representing the heat accumulation capacity per unit time under this exposure time setting. The system calculates a time variable based on the remaining runtime and task type of the current imaging task. This time variable is a thermal safety adjustment weight pre-calculated during the calibration phase according to the remaining runtime of different tasks. The system obtains a specific value by looking up a table; a value less than 1 indicates a need to reduce the thermal accumulation trend. The system calculates the reduced exposure time thermal accumulation coefficient by multiplying the original exposure time thermal accumulation coefficient with this time variable to obtain a smaller reduced exposure time thermal accumulation coefficient. This reduced coefficient is then combined with the corresponding exposure time to form a set of preliminary adjustment parameters. If the system has multiple preset available exposure time levels, the reduced exposure time thermal accumulation coefficient is calculated in the same way for each candidate exposure time, thus obtaining multiple sets of preliminary adjustment parameters. The system then integrates each set of preliminary adjustment parameters with the gain-value current amplification model. This model is established during circuit calibration by injecting a reference current of known amplitude under different gain settings and measuring the output current. The ratio of the output current to the input current is used as the amplification ratio for that gain setting, and a one-to-one correspondence table of gain settings and amplification ratios is created. During runtime, the system assigns a set of candidate gain values ​​to each set of preliminary adjustment parameters. For each candidate gain value, the corresponding amplification ratio is looked up in the amplification ratio table, and then multiplied by the estimated input current under the current task to obtain the corresponding predicted amplified output current value. This expands upon each set of preliminary adjustment parameters to obtain a parameter combination consisting of exposure time, a reduced exposure time heat accumulation coefficient, gain setting, and amplified output current. The humidity impact parameter is directly provided by the interference adjustment coefficient calculated in the previous environmental interference correction step. This coefficient expresses the change in actual heat conduction capacity relative to the standard heat conduction rate under current humidity conditions. The system adds this humidity impact parameter to each parameter combination to uniformly consider the impact of environmental humidity on heat accumulation during compatibility judgment. All these parameter combinations constitute an extended set.The system then performs a compatibility check on each set of parameters in the extended set. The calculation process for the compatibility check is as follows: On the one hand, based on the exposure time, the heat accumulation coefficient of the reduced exposure time, and the humidity effect parameters, the total heat accumulation for the remaining time of the task is calculated. This total heat accumulation is compared with the maximum allowable heat accumulation corresponding to the overheat protection threshold. If the predicted total heat accumulation exceeds the allowable upper limit derived from the overheat protection threshold, the set of parameters is determined to be incompatible with thermal safety and is removed from the extended set. On the other hand, the predicted amplified output current value corresponding to the gain setting is compared with the maximum operating current specified by the circuit design. If the amplified output current exceeds the maximum operating current, the set of parameters is determined to be incompatible with current amplification behavior and is also removed. At the same time, the system compares the deviation of the predicted temperature rise trajectory of the set of parameters with the historical stable temperature rise trajectory based on the stable operating samples recorded in the historical temperature data. When the deviation exceeds the pre-statistically obtained allowable deviation range, it is considered that it may not be able to maintain stable operation for a long time under the environmental humidity conditions, and the set of parameters is also deleted from the set. After performing the above three-category judgments on thermal safety, current amplification, and stability for each parameter combination in the extended set, the system retains only the parameter combinations that simultaneously meet all the judgment conditions. These retained parameter combinations are then sorted according to the temperature margin provided by the distance index and their historical stability performance, arranged from best to worst to form an adjusted parameter candidate set, which serves as the direct input when selecting the final configuration of the imaging system.

[0031] S6 includes extracting a set of adjusted candidate parameters matching the current real-time task from a preset parameter library; fusing external variables through environmental adaptation adjustment, wherein the environmental adaptation adjustment is categorized according to variable type and matched and fused one by one to obtain a preliminary screening set; using the preliminary screening set, performing multi-dimensional stability verification, wherein the multi-dimensional stability verification verifies the fluctuation range of each parameter in the set; fusing hardware load sampling data with the verification results to obtain a synchronous comparison index; judging the relationship between the index and a threshold through the synchronous comparison index; if the index exceeds the preset threshold, activating the dynamic resource allocation process, wherein the dynamic resource allocation process selects supplementary parameters from the backup library to determine extended adaptation parameters; obtaining the extended adaptation parameters, combining them with the real-time task to generate the final output, wherein the fusion integrates task requirements through parameter superposition to obtain the imaging system configuration.

[0032] In this embodiment, records of the same type as the current real-time task are first read from a preset parameter library by task type index. The parameter library is obtained by long-term storage of adjusted parameters that have been screened for thermal safety and stability in the aforementioned steps. Each record contains at least quantitative data such as exposure time, exposure time thermal accumulation coefficient, gain value, humidity influence parameter, historical stability score, and thermal safety margin calculated in step S5. When the current real-time task enters the system, it carries descriptive information such as task type, target resolution, target frame rate, and task duration. The system uses task type as the first screening condition to remove records with inconsistent task types from the parameter library. Then, records with resolution lower than the target resolution or maximum supported frame rate lower than the target frame rate are removed from the remaining records. All the remaining records constitute a set of adjusted parameters that match the current real-time task.The system then performs environmental adaptation adjustments on this set to incorporate external variables. These external variables are the ambient temperature, humidity, wind speed, and power supply fluctuation levels actually collected by the system at the operating site. During the calibration phase, the system categorizes these external variables into two main types based on extensive experimental results: thermal environment and electrical. The thermal environment category includes ambient temperature and wind speed, while the electrical category includes power supply fluctuation levels. Humidity is treated as a separate humidity-affecting parameter. During the environmental adaptation adjustment process, the system first looks up the corresponding thermal influence coefficient in a pre-established thermal environment influence table based on the current ambient temperature and wind speed. This thermal influence coefficient is obtained by measuring the temperature change per unit time under varying ambient temperature and wind speed conditions in a laboratory while keeping the internal load constant. The ratio of the temperature change under different operating conditions to the temperature change under a standard environment is used to represent the amplification or reduction ratio of the heat accumulation rate under the current environment relative to the standard environment. The system multiplies this thermal influence coefficient by the exposure time heat accumulation coefficient in each record to obtain the corrected exposure time heat accumulation coefficient under the current thermal environment. Simultaneously, the system looks up the humidity correction factor in a humidity-affecting parameter table based on the current ambient humidity. This correction factor is obtained during the calibration phase by changing the ambient humidity while maintaining a constant internal load. With other conditions remaining constant, the system compares the ratio of heat transfer per unit time to heat transfer under standard humidity conditions. The system then multiplies the humidity correction factor by the previously corrected exposure time heat accumulation coefficient to obtain the final exposure time heat accumulation coefficient, taking into account the combined effects of ambient temperature, wind speed, and humidity. For electrical external variables, the system looks up the current safety factor in the power supply stability impact table based on the current power supply fluctuation level. This factor, determined during calibration by simulating different power supply fluctuation conditions, recording current spikes, and circuit fault probabilities, is the ratio of the maximum safe current under fault-free conditions to the nominal maximum current. The system multiplies the current safety factor by the amplified output current corresponding to the original gain value in each record to obtain the predicted upper limit of the amplified output current under the current power supply conditions. If the amplified output current in the original record is greater than the predicted value, it is truncated to the predicted value, and the record is marked as having reached the safety limit. After completing the above environmental adaptation adjustment, the system calculates the difference between the corrected maximum heat accumulation value and the overheat protection threshold for each record, and removes records with a difference less than zero. At the same time, records with amplified output current exceeding the absolute maximum allowable current of the circuit are also removed. The remaining records constitute the preliminary screening set.The system then performs multi-dimensional stability checks on the initial screening set. These multi-dimensional checks include temperature fluctuation, current fluctuation, and task smoothness. During the calibration phase, long-term operational tests and fault statistics were used to determine the temperature fluctuation threshold, current fluctuation threshold, and smoothness threshold for each of these three dimensions. The temperature fluctuation threshold is set as the maximum difference between the maximum and minimum temperature values ​​among all samples that did not experience thermal runaway or significant image quality degradation. The current fluctuation threshold is set as the maximum difference between the maximum and minimum current values ​​among samples that did not experience circuit faults. The smoothness threshold is set as the maximum deviation of the consecutive frame time interval among samples that did not experience frame drops or significant stuttering. In the actual verification process, the system uses thermal and current behavior prediction models to simulate and predict the temperature and current changes of each record in the initial screening set during the remaining task time. Within each fixed-length time window, the difference between the maximum and minimum temperature values ​​is calculated as the temperature fluctuation range, and the difference between the maximum and minimum current values ​​is calculated as the current fluctuation range. Simultaneously, based on the same prediction results, the consecutive frame output time interval under this parameter combination is obtained. The maximum deviation is used as an indicator of operational smoothness and compared with the corresponding thresholds to record whether it exceeds the limit. Subsequently, the system integrates these predicted fluctuation results with actual hardware load sampling data. The hardware load sampling data consists of the processing load percentage, storage usage, and total power consumption values ​​measured by the system at a fixed sampling period during the current task's operation. The system calculates the average value of each type of load data within a statistical time window, and then calculates the difference between these average values ​​and the average load benchmark value of the same type of task under historical normal operating conditions. This difference is then divided by the benchmark value to obtain the load deviation percentage. The larger the load deviation percentage, the closer the current system resources are to saturation. Based on the above data, the system calculates a synchronization comparison index for each record. The synchronization comparison index is calculated by weighting and summing the proportions of temperature fluctuation range to temperature fluctuation threshold, current fluctuation range to current fluctuation threshold, operational smoothness to operational smoothness threshold, and the three types of load deviation percentages according to the fixed weights set during the calibration phase. The result is a value that is not less than zero. The larger this value, the higher the overall risk of the parameter combination under the current environment and load conditions. The preset threshold, also known as the synchronization comparison threshold, is determined through a large number of samples during the system calibration phase. Specifically, the system is run under different parameter configurations and various environmental conditions, and the corresponding synchronization comparison indicators are calculated. It is also marked whether thermal protection actions, circuit overloads, or significant frame rate drops occur. The maximum value of the synchronization comparison indicator among all samples that do not have problems is selected as the synchronization comparison threshold and written into the configuration file. During runtime, the system compares the synchronization comparison indicator of each record with this threshold. When the indicator is less than or equal to the threshold, the record is considered to be stable and usable under the current conditions. When the indicator is greater than the threshold, the record is considered to have too high a risk under the current hardware and environmental conditions, and the dynamic resource allocation process needs to be activated.The dynamic resource allocation process selects supplementary parameters from the backup library. The backup library is an additional set of conservative parameter records with lower resource consumption and less heat accumulation, which are stored on top of the parameter library. These records have also undergone safety and stability verification during the calibration phase, but the image effect or performance indicators are slightly conservative. The system selects records from the backup library based on the current hardware load deviation percentage. These records can theoretically significantly reduce temperature fluctuations or current fluctuations without causing the resolution and frame rate to fall below the task requirements. These supplementary records are merged with the records in the initial screening set whose synchronization comparison indicators have not exceeded the limits. The merged set is then re-executed with environmental adaptation adjustment and multi-dimensional stability verification to obtain an extended adaptation parameter set that meets the requirement that the synchronization comparison indicators do not exceed the threshold. Finally, the system retrieves extended adaptation parameters from the extended adaptation parameter set and combines them with the real-time task to generate the final output. Specifically, it sets scoring rules according to the resolution requirements, frame rate requirements, and maximum allowable temperature in the task description, calculates a comprehensive score for each record in the set, and accumulates the comprehensive score by weighting the thermal safety margin, the reverse value of the synchronous comparison index, the historical stability score, and the matching degree with the task target. Then, it selects the record with the highest comprehensive score as the target parameter for this imaging process, and overlays the exposure time, exposure time thermal accumulation coefficient, gain value, and corresponding humidity influence parameters in the record onto the configuration template of the current real-time task in a parameter overlay manner to form an imaging system configuration containing all necessary imaging parameters.

[0033] S7 employs the final adapted imaging system configuration, performs fixed-duration segmented monitoring, and triggers data acquisition in real time through abnormal interruptions to obtain hardware interruption response timing data during continuous operation. For the hardware interruption response timing data, when an anomaly is detected, it backtracks the corresponding sensor node thermal conduction delay and power consumption curve mapping data from the simulation log records. The simulation log records are extracted from a preset system backup to obtain basic parameters for fine-tuning. Based on these basic parameters, image signal calibration variables are fused. These image signal calibration variables are extracted from a preset library and matched item by item to determine an extended stability verification set. The extended stability verification set is then acquired and subjected to multi-dimensional comparisons. These multi-dimensional comparisons verify the fluctuation range of each parameter within the set and fuse hardware load sampling data to determine the overall operational stability index. If the overall operational stability index exceeds a preset threshold, the backup sensor node adjustment is activated, and supplementary parameters are selected from the backup library to obtain an optimized imaging system configuration.

[0034] In this embodiment, the final adapted imaging system configuration obtained in the previous steps is first used. The exposure time, gain value, and various temperature-related optimization parameters determined in the preceding steps are written into the configuration area within the imaging system all at once. According to the fixed-duration monitoring parameters determined in the configuration phase, the entire real-time imaging task's runtime is evenly divided into several consecutive monitoring time periods. The duration of each monitoring time period is determined through multiple experiments during the system design and calibration phase to be a specific value that reflects temperature and interruption changes without causing excessive data volume, and is then fixed as a fixed-duration segmented monitoring parameter. During the actual operation of the imaging system, the system continuously listens for hardware interrupt signals within each monitoring time period and records the number of interrupts within that period, the start and end times of each interrupt response, and the overall temperature of that period at the end of each segment. In addition to power consumption statistics, the system also pre-sets abnormal interrupt judgment thresholds, including a single interrupt response time threshold and an interrupt count threshold within a unit monitoring period. These two thresholds are determined by collecting interrupt response data over a long period of time under normal working conditions, statistically analyzing the maximum interrupt response time and the maximum interrupt count within a unit time under normal conditions, and then adding a safety margin to each threshold. During operation, if the response time of an interrupt exceeds the interrupt response time threshold, or the number of interrupts within a certain monitoring period exceeds the interrupt count threshold, the system immediately marks the interrupt as an abnormal interrupt. Around the trigger time of the abnormal interrupt, within a very short time window, it collects multiple signals, including temperature, current, and power consumption, at a sampling frequency higher than the normal monitoring frequency. This forms hardware interrupt response timing data in chronological order and stores it in the operation record of this task.The system then performs cyclical monitoring of the accumulated hardware interrupt response timing data. When multiple abnormal interrupts occur consecutively within a certain monitoring period, or when the response time of a single abnormal interrupt exceeds the response time limit determined based on fault samples in reliability testing, the system determines that the interrupt behavior under the current configuration is unstable and requires backtracking analysis. Therefore, it retrieves simulation log records from a preset system backup. These simulation log records are structured data sets generated and periodically backed up using a virtual load simulation system during the system development and parameter optimization phases. They contain sensor node thermal conduction delay data and power consumption curve mapping data for different task types and configurations, and are indexed and stored by task type, configuration identifier, and time stamp. During backtracking, the system first selects the data from the backup that matches the current task's task type and the parameter identifiers included in the final adapted imaging system configuration. The system uses a subset of simulation logs corresponding to the configuration. Then, through time alignment, it maps the time of abnormal interruption in the hardware interrupt response timing data to the time in the corresponding simulation log subset when the thermal conduction delay increases sharply or power consumption fluctuates significantly. The system extracts the thermal conduction delay and power consumption mapping values ​​for each node at the corresponding time, based on the sensor node number. To obtain basic indicators that can be used to quantify fine-tuning, the system subtracts these values ​​from the baseline thermal conduction delay and baseline power consumption mapping values ​​recorded in the logs during normal operation. This yields the delay and power consumption deviations of each node relative to the baseline state under the current abnormal state. The average deviation of the same node is then calculated over each monitoring period to form basic indicators for parameter fine-tuning summarized by node. These indicators accurately indicate which nodes have the largest deviations from normal levels in the current configuration and environment, making them suitable as key targets for subsequent parameter fine-tuning.Next, the system fuses image signal calibration variables by fine-tuning basic indicators using parameters. These image signal calibration variables are stored in a preset library, which is established during the image quality calibration phase. Each calibration record includes brightness correction, contrast correction, noise suppression strength, and a corresponding exposure time and gain combination. It also records whether the measured image sharpness and signal-to-noise ratio under this combination reach a preset lower limit. During fusion, the system reads the basic indicators for parameter fine-tuning one by one. Based on the exposure time and gain ranges revealed in the indicators, it searches the preset library for all image signal calibration variables whose exposure time and gain are compatible with those ranges. For each candidate image signal calibration variable, the system replaces the current basic indicator with the exposure time and gain from the variable. The local exposure and gain settings of the corresponding nodes are set according to the basic indicators. The filtering intensity in the signal processing link is adjusted according to the noise suppression intensity in the variables. Then, the known correspondence between the thermal conduction delay and power consumption deviation and the exposure and gain changes in the simulation log is used to recalculate the thermal conduction delay and power consumption deviation of the node under this combination. If the calculation results show that the delay deviation and power consumption deviation are reduced relative to the original basic indicators and the original image sharpness and signal-to-noise ratio of the calibration variable have reached or exceeded the preset lower limit, then the image signal calibration variable and the corresponding node parameter fine-tuning basic indicators are combined into a record and added to the extended stability verification set until all basic indicators are matched with all image signal calibration variables that meet the conditions, forming the extended stability verification set. The system then acquires an extended stability verification set, performs multi-dimensional comparisons and overall operational stability assessments on each record within the set. These multi-dimensional comparisons include temperature, power consumption, and image signal dimensions, and incorporate hardware load sampling data during the calculation process. Specifically, the system utilizes the temperature and power consumption response characteristics tested in the simulation logs, combined with the exposure time, gain, and image signal calibration variables corresponding to each record in the extended stability verification set, to calculate the difference between the maximum and minimum temperatures at each node within a preset observation time window as the temperature fluctuation range for that record. Simultaneously, it calculates the difference between the maximum and minimum power consumption as the power consumption fluctuation range. In the image dimension, by retrieving the corresponding image sharpness and noise levels from a preset calibration library, it derives the maximum change in image quality over different time periods as the image signal fluctuation range. These fluctuation ranges are compared with the temperature fluctuation safety threshold, power consumption fluctuation safety threshold, and image signal fluctuation safety threshold obtained statistically from a large number of stable samples during the long-term reliability testing phase. The proportion of the fluctuation range in each dimension to the safety threshold of that dimension is recorded as the relative temperature fluctuation ratio, relative power consumption fluctuation ratio, and relative image signal fluctuation ratio.Meanwhile, the hardware load sampling data consists of values ​​such as processing load, storage usage, and power output collected by the system at fixed intervals during current operation. The system averages these load sampling values ​​within the same observation time window and subtracts them from the average load value of similar tasks in historical stable samples. The difference is then compared to the historical average to calculate the load deviation percentage. The larger this percentage, the closer the current hardware resources are to their limits. To obtain a single overall operational stability index, the system pre-determines the weighting ratios of temperature, power consumption, and image signal in the calibration phase, as well as the load deviation. These weights are fixed values ​​obtained by analyzing the correlation between the degree of anomaly in each dimension and the occurrence of faults in different fault samples. During operation, the system weights the aforementioned relative temperature fluctuation ratio, relative power consumption fluctuation ratio, relative image signal fluctuation ratio, and load deviation percentage according to these weights to obtain an overall operational stability index value that is not less than zero. The larger this value, the higher the overall instability risk. The preset threshold, or overall operational stability threshold, was determined through a large number of samples during the system development and verification phase. Specifically, the imaging system was run under various parameter combinations and different environmental conditions. The overall operational stability index corresponding to each combination was recorded, and it was noted whether the combination experienced hardware overheating protection, continuous abnormal interruptions, or significant degradation of image quality during long-term operation. Then, the maximum value of the overall operational stability index was taken from all combinations that did not exhibit the above problems, and this value was fixed into the configuration file as the overall operational stability threshold.During actual operation, the system calculates the overall operational stability index for each record in the extended stability verification set and compares it with a preset threshold. If the index is less than or equal to the preset threshold, the parameter combination corresponding to that record is considered acceptable under the current environment and load. If the overall operational stability index is greater than the preset threshold, it indicates that the parameter combination corresponding to that record has an unacceptable stability risk under the current conditions, triggering the backup sensor node adjustment process. The backup sensor node adjustment relies on the backup library, which is established during the system design and sensor array deployment. For each physical area, a set of backup sensor nodes is pre-selected and configured with relatively conservative exposure and gain parameters. These parameters have been verified during the calibration phase to have low heat accumulation and small power consumption fluctuations. The system determines the transmission frequency of abnormally concentrated events based on hardware interrupt response timing data and simulation log backtracking results. For the sensor node region, select backup sensor nodes that are physically adjacent to or cover the same field of view as the abnormal region from the backup library, and read the supplementary parameters corresponding to these nodes. Use these supplementary parameters to replace the exposure time and gain parameters used by some or all sensor nodes in the original abnormal region. At the same time, update the corresponding thermal conduction delay and power consumption curve mapping data to the parameter fine-tuning basic index. Then repeat the aforementioned extended stability verification and overall operational stability index calculation process for the replaced combination. From all parameter combinations whose overall operational stability index does not exceed the preset threshold, select the best combination according to the comprehensive ranking principle of large thermal safety margin, good image quality maintenance, and small overall operational stability index. Add the exposure time, gain value, and corresponding image signal calibration variables in this combination to the original final adapted imaging system configuration to form the optimized imaging system configuration.

[0035] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing imaging parameters for visual imaging, characterized in that, include: S1. By receiving the parameter configuration output by the optimization algorithm, the exposure time and gain value are extracted and injected into the virtual load simulation system using a dynamic parameter loading method. The preset ambient temperature gradient field and chip power consumption curve mapping are loaded to obtain the preliminary simulation running state including the thermal conduction delay of the sensor node; S2. Based on the preliminary simulation operation status, set fixed-duration segmented monitoring and high-temperature stage data acquisition encryption rules, and simultaneously enable adaptive acquisition interval adjustment and acquisition frequency upper limit lock. In the virtual load simulation system, simultaneously start hardware interrupt response timing simulation and exposure time thermal accumulation coefficient calculation to determine the simulation operation log record under multi-scene illumination mutation injection. S3. Based on the simulation operation log, extract the data packets that are triggered by abnormal interruption and collect them immediately. Combine the sensor node distribution information and timestamp alignment method, collect encrypted real-time temperature data streams at the stage when the temperature is higher than the preset threshold, and determine the initial trend of temperature monitoring. S4. By monitoring the initial trend of temperature, the peak fluctuation is analyzed using polynomial curve fitting technology. If the peak fluctuation exceeds the preset threshold alarm range, the environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve is triggered to obtain the corrected temperature change data.

2. The imaging parameter optimization method for visual imaging according to claim 1, characterized in that: S1 includes: By receiving the parameter configuration output by the preset optimization algorithm, the exposure time value and gain value are extracted and injected into the virtual load simulation system. The preset ambient temperature gradient field is loaded to calculate the temperature distribution and obtain the temperature distribution matrix. For the temperature distribution matrix, the power consumption distribution of the sensor node is mapped using the chip power consumption curve to obtain the heat source intensity value at the sensor node and determine the heat conduction path; If the heat conduction path exceeds the preset threshold, noise interference simulation is injected to adjust the heat source distribution by superimposing random noise signals, obtain the corrected heat source intensity value, and determine the heat conduction delay time. Based on the heat conduction delay time, the gain value is dynamically adjusted and injected. A real-time calibration mechanism is applied to compensate for parameters by comparing the delay deviation, thus obtaining simulation environment data that includes delay compensation. By integrating simulation environment data with preset preliminary simulation operation states, the final index of thermal conduction delay of sensor nodes is obtained.

3. The imaging parameter optimization method for visual imaging according to claim 1, characterized in that: S2 includes: Based on the preliminary simulation operation status, fixed-duration segmented monitoring data and high-temperature stage acquisition encryption rule parameters are obtained. An adaptive acquisition interval adjustment mechanism is used to fuse the upper limit lock value of the acquisition frequency to obtain the preliminary monitoring sequence. For the initial monitoring sequence, noise interference compensation service is injected into the virtual load simulation system to obtain the sequence distribution after compensation and determine the hardware interrupt response timing simulation start point. The heat accumulation coefficient calculation process of exposure time is loaded through the compensated sequence distribution. The calculation process uses the integral method to accumulate and sum the heat changes in the time period in the sequence distribution. The upper limit of the integral is the lower limit of the sequence duration as the starting point, and the heat accumulation distribution matrix is ​​obtained. If the heat accumulation distribution matrix exceeds the preset threshold, inject multiple scene illumination mutations, obtain the simulation environment data after mutation injection, and determine the trigger conditions for running log recording. Based on the trigger conditions recorded in the operation log, the latency compensation injection service is integrated, and the final simulation operation log is obtained by calibrating and adjusting the time latency deviation.

4. The imaging parameter optimization method for visual imaging according to claim 1, characterized in that: S3 includes: Extract the instantaneous data packets generated by the abnormal interruption from the simulation operation log records, and obtain the encrypted frequency sampling sequence and corresponding trigger time lock information carried by the instantaneous data packets; Based on the pre-established sensor distribution map and node location mapping, the instantaneous data packets are timestamped to obtain the correspondence between the spatial location locked by each sensor node at the trigger time and the encrypted frequency sampling sequence. For stages where the temperature exceeds a preset threshold, the corresponding encrypted frequency sampling sequence is filtered by locking the trigger time information, and the real-time temperature data stream of each node is obtained by decryption; The real-time temperature data stream is spatially mapped according to the sensor distribution map to determine the correlation matrix between the temperature value of each node and the node location. Based on the correlation matrix, an initial trend line of the temperature data stream is plotted on the sensor distribution map to determine the initial trend of temperature monitoring.

5. The imaging parameter optimization method for visual imaging according to claim 1, characterized in that: S4 includes: By monitoring the initial trend of temperature, the peak fluctuation sequence is obtained by using polynomial curve fitting technology. The polynomial curve fitting technology takes the trend data points as input, calculates the polynomial coefficients by the least squares method, and outputs the peak fluctuation sequence. For a peak fluctuation sequence, the fluctuation amplitude is analyzed. The analysis is performed by calculating the maximum absolute value of the difference between adjacent points in the sequence to determine the peak fluctuation value. If the peak fluctuation value exceeds the preset threshold critical alarm range, an environmental interference correction mechanism based on the air-cooled heat dissipation attenuation curve is triggered. The air-cooled heat dissipation attenuation curve adjusts the heat conduction parameters by fusing wind speed data. The fusing multiplies the wind speed data by an attenuation factor and then superimposes it into the curve equation. Based on the environmental interference correction mechanism, the humidity-related interference component is extracted from the mechanism to obtain the humidity interference factor and the interference adjustment coefficient is obtained. The interference adjustment coefficient is calculated by dividing the humidity interference factor by the standard thermal conductivity. Temperature change data is processed using an interference adjustment factor. The corrected temperature change data is obtained by multiplying the temperature change data by the interference adjustment factor.

6. The imaging parameter optimization method for visual imaging according to claim 1, characterized in that, This also includes S5, which compares the corrected temperature change data with historical temperature data to the distance from the overheat protection threshold. If the distance is less than the preset critical alarm standard, the alternative parameter switching process is activated to determine the adjusted parameter alternative set, which includes reducing the heat accumulation coefficient of exposure time and the current amplification model of gain value. Specifically, this includes: Based on the corrected temperature data and combined with historical temperature data, the distance between it and the overheat protection threshold is calculated. The distance is obtained by subtraction to get the difference between the current value and the threshold, thus obtaining the distance index. For the distance indicator, if the distance indicator is less than the preset critical alarm standard, the alternative parameter switching process is activated. The heat accumulation coefficient of reducing exposure time is extracted from the process. The heat accumulation coefficient of reducing exposure time is adjusted by multiplying and fusing time variables to determine the initial adjustment parameters. By initially adjusting the parameters and integrating the gain-value current amplification model, the gain-value current amplification model generates an amplified output by scaling the current input proportionally, thereby obtaining an extended set of parameters that include the influence of humidity. By expanding the set, the compatibility between parameters in the set is determined, and an adjusted set of alternative parameters is obtained.

7. The imaging parameter optimization method for visual imaging according to claim 6, characterized in that, It also includes S6, which extracts a set of adjusted parameters matching the current real-time task from a preset parameter library, and obtains the final adapted imaging system configuration through multi-dimensional stability verification and hardware load sampling synchronous comparison, specifically including: Extract a set of adjusted parameters that match the current real-time task from the preset parameter library, and integrate external variables through environmental adaptation adjustment. The environmental adaptation adjustment is categorized according to variable type and matched and integrated one by one to obtain a preliminary screening set. A preliminary screening set is used to perform multi-dimensional stability verification. The multi-dimensional stability verification verifies the fluctuation range of each parameter in the set. The verification results are then integrated with hardware load sampling data to obtain synchronous comparison indicators.

8. The imaging parameter optimization method for visual imaging according to claim 7, characterized in that: S6 further includes: By synchronously comparing indicators, the relationship between indicators and thresholds is determined. If an indicator exceeds a preset threshold, the dynamic resource allocation process is activated. The dynamic resource allocation process selects supplementary parameters from the backup library and determines the extended adaptation parameters. The extended adaptation parameters are obtained and combined with real-time task fusion to generate the final output. The fusion integrates task requirements through parameter superposition to obtain the imaging system configuration.

9. The imaging parameter optimization method for visual imaging according to claim 7, characterized in that, It also includes S7, using the final adapted imaging system configuration, continuing to perform fixed-duration segmented monitoring and real-time trigger acquisition for abnormal interruptions. If a hardware interruption response timing abnormality is detected again during continuous operation, the corresponding sensor node thermal conduction delay and power consumption curve mapping data in the simulation log record are retrieved, parameters are adjusted, and the overall operational stability is judged. Specifically, this includes: Using the final adapted imaging system configuration, fixed-duration segmented monitoring is performed, and data acquisition is triggered immediately by abnormal interruption to obtain the timing data of hardware interrupt response during continuous operation; For hardware interrupt response timing data, when an anomaly is detected, the corresponding sensor node thermal conduction delay and power consumption curve mapping data in the simulation log record are traced back. The simulation log record is extracted from the preset system backup to obtain the basic indicators for parameter fine-tuning. By fine-tuning the basic indicators through parameters and integrating image signal calibration variables, the image signal calibration variables are extracted from a preset library and matched with indicator values ​​one by one to determine the extended stability verification set.

10. The imaging parameter optimization method for visual imaging according to claim 9, characterized in that: The S7 also includes: Obtain an extended stability verification set and perform multi-dimensional comparisons. The multi-dimensional comparisons verify the fluctuation range of each parameter in the set and integrate hardware load sampling data to determine the overall operational stability indicators. If the overall operational stability index exceeds the preset threshold, the backup sensor node adjustment is activated, and supplementary parameters are selected from the backup library to obtain the optimized imaging system configuration.

Images