A machine learning-based illumination feedback method, apparatus, device, and medium

Abstract

This invention relates to the technical field of illumination environment state perception and feedback control, and in particular to a machine learning-based illumination feedback method, apparatus, device, and medium. The method includes inputting illumination intensity data and attitude data into a trained data validity discrimination model to obtain validity label information; filtering out invalid data from the illumination intensity data based on the validity label information to obtain valid illumination data; performing anomaly detection processing based on the illumination feature information and a preset illumination health threshold range to obtain illumination anomaly detection result information; sending the illumination anomaly warning information to a terminal; receiving parameter configuration instructions returned by the terminal; and updating the monitoring parameters at the acquisition end based on the parameter configuration instructions. This invention effectively solves the problems of insufficient reliability and poor feedback adaptability in existing illumination monitoring systems.

Description

Technical Field

[0001] This invention relates to the technical field of illumination environment state perception and feedback control, and in particular to an illumination feedback method, device, equipment and medium based on machine learning. Background Technology

[0002] Children aged 6-12 are in a critical period of visual development. In daily learning and life scenarios, factors such as the switching of indoor and outdoor light sources, the use of display screens, and the differences in lighting conditions between classrooms and homes cause the lighting environment to be highly dynamic. Long-term deviations or frequent fluctuations in light levels may be related to the risk of myopia development. Therefore, there is a real need to record and warn children about the risks of their lighting environment over a long period of time.

[0003] In existing technologies, most light monitoring products for homes or schools mainly display "illuminance values," typically relying on single sampling or short-cycle statistical results for alerts. This makes it difficult to reliably reflect the true light exposure situation. On the one hand, children often move, block, turn their heads, or tilt their heads while wearing or carrying monitoring devices, causing instantaneous changes, spikes, and discontinuities in the collected illuminance data. Simply relying on numerical values ​​for judgment can easily lead to false triggers or missed triggers. On the other hand, light changes in different scenarios have a cumulative effect over time, and instantaneous values ​​alone cannot characterize the light exposure pattern, resulting in a lack of interpretability and actionability in the alert content. Summary of the Invention

[0004] To address the issues of insufficient reliability and poor feedback adaptability in existing illumination monitoring systems, this application provides an illumination feedback method, apparatus, device, and medium based on machine learning.

[0005] The above-mentioned objective of this application is achieved through the following technical solution: A machine learning-based illumination feedback method, comprising: Acquire light intensity data collected by the acquisition terminal, and acquire attitude data corresponding to the time of the light intensity data; The illumination intensity data and the attitude data are input into a trained data validity discrimination model to obtain validity labeling information; Based on the validity labeling information, invalid data is filtered out from the light intensity data to obtain valid light data; Perform feature extraction processing on the effective illumination data to obtain illumination feature information; Based on the light feature information and the preset light health threshold range, anomaly detection processing is performed to obtain light anomaly detection result information; If the lighting anomaly determination result indicates an anomaly, continuous determination processing is performed on the lighting anomaly determination result to generate lighting anomaly warning information; The abnormal light warning information is sent to the terminal, the parameter configuration instruction information returned by the terminal is received, and the monitoring parameters of the acquisition terminal are updated based on the parameter configuration instruction information.

[0006] By adopting the above technical solution, the light intensity data collected by the acquisition end and the posture data corresponding to the light intensity data time can be used together to participate in the data validity judgment. In the scenario of light intensity data fluctuation caused by posture changes, the light intensity data is marked as valid and invalid data is filtered out accordingly, thereby forming more stable valid light intensity data and supporting subsequent feature extraction and anomaly detection processing. It can extract light feature information from valid light intensity data and compare it with the preset light health threshold range to generate light anomaly detection result information. When the light anomaly detection result information indicates an anomaly, it combines continuous judgment processing to generate light anomaly warning information to reduce the inconsistency of anomaly prompts caused by instantaneous fluctuations. It can send the light anomaly warning information to the terminal and receive the parameter configuration instruction information returned by the terminal to update the monitoring parameters of the acquisition end, thereby realizing closed-loop linkage configuration between the monitoring side and the terminal side and keeping the monitoring parameters consistent with the actual use scenario.

[0007] In a preferred embodiment, this application can be further configured as follows: inputting the illumination intensity data and the pose data into a trained data validity discrimination model to obtain validity labeling information includes: The light intensity data is processed by performing a rate of change calculation within a preset time window to obtain light change characteristic information, and the attitude data is processed by performing a stability calculation within the preset time window to obtain attitude stability characteristic information. The illumination change feature information and the attitude stability feature information are input into the trained data validity discrimination model to obtain validity label information corresponding to the preset time window.

[0008] By adopting the above technical solution, the change rate calculation of light intensity data can be performed within a preset time window to form light change feature information characterizing the amplitude and speed of light fluctuations. Simultaneously, stability calculation of attitude data can be performed within the same preset time window to form attitude stability feature information characterizing the degree of attitude change. This ensures that the light change feature information and attitude stability feature information are consistent in the time dimension and can jointly characterize the dynamic state of the data acquisition process. The light change feature information and attitude stability feature information can be input into a trained data validity discrimination model to output validity label information corresponding to the preset time window. This achieves time window-based validity discrimination and establishes a correspondence between validity label information and light intensity data at the window level. This facilitates subsequent filtering of light intensity data according to the validity label information and maintains the repeatability and consistency of the data validity discrimination process.

[0009] In a preferred embodiment, this application can be further configured such that: the step of filtering out invalid data from the illumination intensity data based on the validity labeling information to obtain valid illumination data includes: The light intensity data is traversed in the order of sampling time according to the light intensity data, and the validity marker information corresponding to the sampling time is read. If the validity marker information indicates that the sample is valid, the light intensity data corresponding to the sampling time will be determined as a valid sample value and retained. If the validity marker information indicates invalidity, the light intensity data corresponding to the sampling time is determined as invalid data and filtered out. The remaining valid sampled values ​​are collected in the order of the sampling time to form the valid light data.

[0010] By adopting the above technical solution, the validity marker information corresponding to the sampling time can be read point by point along the sampling time sequence of the light intensity data. When the validity marker information indicates that it is valid, the corresponding light intensity data is retained as a valid sample value. When the validity marker information indicates that it is invalid, the corresponding light intensity data is determined as invalid data and filtered out. Thus, the retention and filtering of light intensity data are directly constrained by the validity marker information and a clear filtering boundary is formed. The retained valid sample values ​​can be collected in the order of sampling time to form valid light intensity data. This avoids the inclusion of abrupt sampling points caused by attitude changes or acquisition anomalies in the valid light intensity data, and makes the data sequence based on which the subsequent feature extraction and anomaly judgment of the valid light intensity data are performed more stable and consistent.

[0011] In a preferred embodiment, this application can be further configured such that: the feature extraction processing performed on the effective illumination data to obtain illumination feature information includes: The effective illumination data is segmented according to the preset time window information to obtain the time window sequence information; In the time window sequence information, cumulative statistical processing is performed on the effective illumination data within each preset time window information to obtain effective illumination duration information; Within the preset time window, the effective illumination data is subjected to mean calculation to obtain average illumination intensity information; Within the preset time window, threshold comparison statistical processing is performed on the effective illumination data to obtain strong light exposure frequency information and weak light exposure frequency information. The strong light exposure frequency information is the statistical result of the number of times the effective illumination data is greater than the second threshold information, and the weak light exposure frequency information is the statistical result of the number of times the effective illumination data is less than the first threshold information. The effective illumination duration information, the average illumination intensity information, the strong light exposure frequency information, and the weak light exposure frequency information are combined to form the illumination characteristic information.

[0012] By adopting the above technical solution, effective illumination data can be segmented according to preset time window information, enabling illumination analysis to have a unified time granularity and alignable statistical boundaries. Within each preset time window, cumulative statistical processing can be performed on the effective illumination data to obtain effective illumination duration information, characterizing the duration of effective sampling within the window. Mean calculation processing can be performed on the effective illumination data within the preset time window to obtain average illumination intensity information, characterizing the overall illumination level within the window. Threshold comparison statistical processing can be performed on the effective illumination data within the preset time window to obtain strong light exposure frequency information and weak light exposure frequency information, characterizing the dispersion and frequency of strong and weak light occurrences. Effective illumination duration information, average illumination intensity information, strong light exposure frequency information, and weak light exposure frequency information can be aggregated to form illumination feature information, allowing subsequent anomaly detection processing to simultaneously perform consistent judgments based on duration, mean, and exposure frequency while maintaining traceable window-level feature records.

[0013] In a preferred embodiment, this application can be further configured as follows: the step of performing mean calculation processing on the effective illumination data within the preset time window information to obtain average illumination intensity information includes: Obtain valid illumination data arranged in the order of sampling time within the preset time window, and extract the sampling time information and illumination intensity sampling value information from the valid illumination data; The sampling interval information is calculated based on the adjacent sampling time information, and the sampling interval information is bound with the light intensity sampling value information corresponding to the adjacent sampling time information to obtain the time-weighted entry information; The time-weighted entry information is processed by calculating the rate of change of illumination to obtain the corrected sampling interval information; Based on the corrected sampling interval information, time-weighted accumulation processing is performed on the light intensity sample value information to obtain weighted accumulation value information, and accumulation processing is performed on the corrected sampling interval information to obtain interval accumulation value information; The average light intensity information is obtained by performing a ratio calculation process using the weighted cumulative value information and the interval cumulative value information.

[0014] By adopting the above technical solution, sampling time information and light intensity sampling value information can be extracted within a preset time window. Sampling interval information can be calculated based on adjacent sampling time information to form time-weighted entry information, thus establishing a clear binding relationship between light intensity sampling value information and the time dimension. Light change rate calculation processing can be performed on the time-weighted entry information to obtain corrected sampling interval information, so that the sampling interval information is adjusted with the light change amplitude and forms a time weight that better matches the light change state within the window. Time-weighted accumulation processing can be performed on the light intensity sampling value information based on the corrected sampling interval information to obtain weighted accumulation value information and interval accumulation value information simultaneously. Thus, even when the sampling interval is uneven or the change within the window is sudden, the mean calculation caliber can still be kept consistent. The average light intensity information can be calculated by the ratio of weighted accumulation value information to interval accumulation information, so that the average light intensity information corresponds to the distribution and change state of sampling time and facilitates connection with subsequent threshold interval determination.

[0015] In a preferred embodiment, this application can be further configured as follows: the anomaly determination process based on the illumination feature information and a preset illumination health threshold range, to obtain illumination anomaly determination result information, includes: Read the preset light health threshold range and parse the preset light health threshold range into first threshold information and second threshold information; The average light intensity information, strong light exposure frequency information, and weak light exposure frequency information are extracted from the light characteristic information. The average light intensity information is compared with the first threshold information to generate weak light anomaly marker information when the average light intensity information is less than the first threshold information. The average light intensity information is compared with the second threshold information to generate strong light anomaly marker information when the average light intensity information is greater than the second threshold information. When at least one of the strong light anomaly marker information and the weak light anomaly marker information exists, the strong light exposure frequency information and the weak light exposure frequency information are subjected to anomaly consistency verification processing, so as to generate the light anomaly judgment result information when the strong light exposure frequency information or the weak light exposure frequency information meets the preset exposure conditions.

[0016] By adopting the above technical solution, the preset light health threshold range can be read and parsed into first threshold information and second threshold information, so that the anomaly judgment has clear interval boundaries and configurable judgment criteria. It can extract average light intensity information, strong light exposure frequency information and weak light exposure frequency information from light feature information, so that the judgment input simultaneously covers two dimensions: overall level and exposure frequency. It can compare the average light intensity information with the first threshold information and the second threshold information respectively to generate weak light anomaly label information and strong light anomaly label information, thereby completing the initial screening of anomaly types based on the mean dimension. When at least one anomaly label information exists, it can further perform anomaly consistency verification processing on the strong light exposure frequency information and the weak light exposure frequency information, and generate light anomaly judgment result information when the preset exposure conditions are met. This ensures that the anomaly judgment simultaneously meets the intensity over-limit and exposure frequency constraints and maintains the consistency of the judgment logic, which is conducive to the connection with the subsequent continuous judgment and early warning generation process.

[0017] In a preferred embodiment, this application can be further configured as follows: when the illumination anomaly determination result information indicates an anomaly, performing continuous determination processing on the illumination anomaly determination result information to generate illumination anomaly early warning information includes: When the lighting anomaly determination result indicates an anomaly, obtain the preset duration information; The continuity determination window information is determined according to the preset duration information; Within the sampling time range corresponding to the continuity determination window information, the illumination anomaly determination result information is read sequentially along the sampling time. Within the sampling time range corresponding to the continuity determination window information, the anomaly retention duration information is statistically analyzed; When the duration of the abnormality reaches the preset duration, the light abnormality warning information is generated.

[0018] By adopting the above technical solution, when the light anomaly judgment result indicates an anomaly, a preset duration information can be obtained, and the continuous judgment window information can be determined accordingly. This gives the warning trigger a clear time constraint and avoids triggering the warning directly based on a single anomaly judgment result. The light anomaly judgment result information can be read sequentially along the sampling time range corresponding to the continuous judgment window information, and the anomaly holding duration information can be statistically analyzed. This makes the warning triggering based on the continuity of the anomaly state within the window rather than discrete point judgments. The light anomaly warning information can be generated when the anomaly holding duration information reaches the preset duration information. This establishes a correspondence between the light anomaly warning information and the continuous anomaly state and keeps the triggering conditions traceable. This facilitates the connection with the subsequent process of sending and receiving parameter configuration instructions to the terminal, thus ensuring that the warning generation and terminal interaction update process have a consistent judgment benchmark and time window boundary.

[0019] The second objective of this invention is achieved through the following technical solution: A machine learning-based illumination feedback device, the machine learning-based illumination feedback device comprising: The data acquisition module is used to acquire light intensity data collected by the acquisition terminal and to acquire attitude data corresponding to the time of the light intensity data. The validity discrimination module is used to input the illumination intensity data and the pose data into a trained data validity discrimination model to obtain validity labeling information; An invalid data filtering module is used to filter out invalid data from the light intensity data based on the validity labeling information to obtain valid light intensity data; The feature extraction module is used to perform feature extraction processing on the effective illumination data to obtain illumination feature information; The anomaly detection module is used to perform anomaly detection processing based on the illumination feature information and the preset illumination health threshold range to obtain illumination anomaly detection result information; The continuous early warning module is used to perform continuous judgment processing on the light anomaly judgment result information to generate light anomaly early warning information when the light anomaly judgment result information indicates an anomaly. The parameter update module is used to send the abnormal light warning information to the terminal, receive the parameter configuration instruction information returned by the terminal, and update the monitoring parameters of the acquisition terminal based on the parameter configuration instruction information.

[0020] By adopting the above technical solution, the light intensity data collected by the acquisition end and the posture data corresponding to the light intensity data time can be used together to participate in the data validity judgment. In the scenario of light intensity data fluctuation caused by posture changes, the light intensity data is marked as valid and invalid data is filtered out accordingly, thereby forming more stable valid light intensity data and supporting subsequent feature extraction and anomaly detection processing. It can extract light feature information from valid light intensity data and compare it with the preset light health threshold range to generate light anomaly detection result information. When the light anomaly detection result information indicates an anomaly, it combines continuous judgment processing to generate light anomaly warning information to reduce the inconsistency of anomaly prompts caused by instantaneous fluctuations. It can send the light anomaly warning information to the terminal and receive the parameter configuration instruction information returned by the terminal to update the monitoring parameters of the acquisition end, thereby realizing closed-loop linkage configuration between the monitoring side and the terminal side and keeping the monitoring parameters consistent with the actual use scenario.

[0021] The above-mentioned objective three of this application is achieved through the following technical solution: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the aforementioned machine learning-based illumination feedback method.

[0022] The fourth objective of this application is achieved through the following technical solution: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the aforementioned machine learning-based illumination feedback method.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. It can combine the light intensity data collected by the acquisition terminal with the posture data corresponding to the light intensity data time to participate in the data validity judgment. In the case of light intensity data fluctuation caused by posture changes, it can mark the validity of light intensity data and filter out invalid data accordingly, thereby forming more stable valid light intensity data and supporting subsequent feature extraction and anomaly detection processing. It can extract light feature information from valid light intensity data and compare it with the preset light health threshold range to generate light anomaly detection result information. When the light anomaly detection result information indicates an anomaly, it can combine continuous judgment processing to generate light anomaly warning information to reduce the inconsistency of anomaly prompts caused by instantaneous fluctuations. It can send light anomaly warning information to the terminal and receive parameter configuration instructions returned by the terminal to update the monitoring parameters of the acquisition terminal, thereby realizing closed-loop linkage configuration between the monitoring side and the terminal side and keeping the monitoring parameters consistent with the actual use scenario. Attached Figure Description

[0024] Figure 1 This is a flowchart of a machine learning-based illumination feedback method in one embodiment of this application.

[0025] Figure 2 This is a schematic diagram of a machine learning-based illumination feedback device according to one embodiment of this application. Detailed Implementation

[0026] The present application will be further described in detail below with reference to the accompanying drawings.

[0027] In one embodiment, such as Figure 1 As shown, this application discloses a machine learning-based illumination feedback method, which specifically includes the following steps: S10: Acquire the light intensity data collected by the acquisition terminal, and acquire the attitude data corresponding to the time of the light intensity data.

[0028] In this embodiment, the acquisition end refers to a wearable or portable acquisition device that integrates a light acquisition device and an attitude acquisition device. The light intensity data refers to the sequence of light intensity sampled values ​​formed by the acquisition end after acquiring the ambient light at the sampling time. The attitude data refers to the sequence of attitude sampled values ​​formed by the acquisition end after acquiring the spatial attitude of the acquisition end at the sampling time. The attitude data corresponding to the light intensity data time refers to the attitude sampled value in the attitude data that corresponds one-to-one with the sampling time of the light intensity data.

[0029] Specifically, the acquisition end samples and reads the ambient light according to a preset sampling frequency to form light intensity sampling value information, and simultaneously records the sampling time information to form light intensity data. The acquisition end samples and reads the spatial attitude at the time corresponding to the sampling time information to form attitude sampling value information, and simultaneously records the sampling time information corresponding to the attitude sampling value information to form attitude data.

[0030] S20: Input the illumination intensity data and attitude data into the trained data validity discrimination model to obtain validity labeling information.

[0031] In this embodiment, the trained data validity discrimination model refers to a discrimination model that has been pre-trained and is able to perform validity discrimination processing on the input light intensity data and attitude data. The validity label information refers to the label result sequence output by the trained data validity discrimination model, which is used to characterize whether the sampled values ​​in the light intensity data belong to invalid interference data. Invalid interference data refers to abnormal sampled values ​​caused by instantaneous light and shadow occlusion and equipment jitter.

[0032] Specifically, the sampling time information and light intensity sample value information in the illumination intensity data are read, and the sampling time information and attitude sample value information in the attitude data are also read. Alignment processing is performed on the sampling time information of the illumination intensity data and the sampling time information of the attitude data to form a time alignment result. When the time alignment result indicates that the alignment is complete, the illumination intensity sample value information and attitude sample value information associated with the same sampling time information are combined according to the input order of the trained data validity discrimination model to form discrimination input data, which is then input into the trained data validity discrimination model. The trained data validity discrimination model performs validity discrimination processing on the discrimination input data and outputs validity label values. The validity label values ​​corresponding to each sampling time information are collected in the sampling time order to form validity label information.

[0033] S30: Based on the validity labeling information, invalid data is filtered out from the light intensity data to obtain valid light data.

[0034] In this embodiment, effective illumination data refers to the sequence of illumination intensity sample values ​​retained after removing invalid interference data from the illumination intensity data.

[0035] Specifically, the validity marker information is read, along with the sampling time information and the light intensity sample value information from the light intensity data. The light intensity sample value information is traversed according to the order of the sampling time information. When the sampling time information is encountered, the validity marker information corresponding to the sampling time information is read. If the validity marker information indicates that the data is valid, the light intensity sample value information corresponding to the sampling time information is written into the valid sample value set. If the validity marker information indicates that the data is invalid, the light intensity sample value information corresponding to the sampling time information is identified as invalid interference data and discarded. After the traversal is completed, the valid sample value set is aggregated according to the order of the sampling time information to form valid light intensity data.

[0036] S40: Perform feature extraction processing on the effective illumination data to obtain illumination feature information.

[0037] In this embodiment, illumination feature information refers to a set of features used to characterize the illumination exposure state after statistical analysis and calculation of effective illumination data.

[0038] Specifically, the sampling time information and light intensity sampling value information in the effective illumination data are read, and a preset time window information is determined based on the sampling time information. The effective illumination data is segmented according to the preset time window information to form a time window sequence information. In the time window sequence information, the adjacent difference calculation is performed on the sampling time information within each preset time window information to obtain the sampling interval information. When the light intensity sampling value information meets the preset threshold comparison condition, the corresponding sampling interval information is cumulatively statistically analyzed to form the effective illumination duration information. The light intensity sampling value information within the preset time window information is time-weighted averaged to form the average light intensity information. The light intensity sampling value information within the preset time window information is compared and statistically analyzed with the first threshold information and the second threshold information in the preset light health threshold interval to form the weak light exposure frequency information and the strong light exposure frequency information. Subsequently, the effective illumination duration information, average light intensity information, strong light exposure frequency information, and weak light exposure frequency information are aggregated according to the preset field order to form illumination feature information.

[0039] S40: Perform anomaly detection processing based on illumination feature information and preset illumination health threshold range to obtain illumination anomaly detection result information.

[0040] In this embodiment, the preset light health threshold range refers to the threshold range used to limit the light to be within the healthy range. The preset light health threshold range is composed of first threshold information and second threshold information. The preset exposure condition refers to the number threshold condition or percentage threshold condition used when judging the frequency information of strong light exposure and weak light exposure.

[0041] Specifically, the system reads a preset light health threshold range and parses it to obtain first and second threshold information. It extracts average light intensity information, strong light exposure frequency information, and weak light exposure frequency information from the light characteristic information. It compares the average light intensity information with the first threshold information and generates a weak light anomaly marker when the average light intensity information is less than the first threshold information. It compares the average light intensity information with the second threshold information and generates a strong light anomaly marker when the average light intensity information is greater than the second threshold information. When either the weak light anomaly marker information or the strong light anomaly marker information exists, it performs exposure condition determination processing on the weak light exposure frequency information and the strong light exposure frequency information and matches them with preset exposure conditions. Finally, it generates a light anomaly determination result information when either the weak light exposure frequency information or the strong light exposure frequency information meets the preset exposure conditions.

[0042] S50: Perform anomaly detection processing based on illumination feature information and preset illumination health threshold range to obtain illumination anomaly detection result information.

[0043] In this embodiment, the light anomaly determination result information refers to the anomaly determination output result formed after threshold comparison and exposure condition verification of light feature information.

[0044] Specifically, the system reads illumination characteristic information and extracts average illumination intensity information, strong light exposure frequency information, and weak light exposure frequency information. It reads a preset illumination health threshold range and parses it to obtain first threshold information and second threshold information. It compares the average illumination intensity information with the first threshold information and generates a weak light anomaly marker when the average illumination intensity information is less than the first threshold information. It compares the average illumination intensity information with the second threshold information and generates a strong light anomaly marker when the average illumination intensity information is greater than the second threshold information. If either the weak light anomaly marker information or the strong light anomaly marker information exists, it reads preset exposure conditions and performs exposure condition matching processing on the weak light exposure frequency information and the strong light exposure frequency information. If either the weak light exposure frequency information or the strong light exposure frequency information meets the preset exposure conditions, it generates an illumination anomaly judgment result. If neither the weak light anomaly marker information nor the strong light anomaly marker information exists, or if neither the weak light exposure frequency information nor the strong light exposure frequency information meets the preset exposure conditions, it generates a non-anomaly judgment marker and writes it into the illumination anomaly judgment result information.

[0045] S60: If the lighting anomaly determination result indicates an anomaly, perform continuous determination processing on the lighting anomaly determination result to generate lighting anomaly warning information.

[0046] In this embodiment, the preset duration information refers to the duration parameter used to limit the minimum duration for which the anomaly needs to be maintained, the continuity determination window information refers to the time range for performing continuity determination based on the preset duration information, and the anomaly maintenance duration information refers to the cumulative duration for which the illumination anomaly determination result information continuously indicates the anomaly.

[0047] Specifically, when the illumination anomaly determination result indicates an anomaly, a preset duration information is read and a continuity determination window information is determined based on the preset duration information. Illumination anomaly determination result information corresponding to the continuity determination window information is read sequentially along the sampling time, and segmented identification processing is performed on the sampling time interval that continuously indicates anomaly. In the segmented identification processing, the difference is calculated based on the adjacent sampling time information to obtain the sampling interval information, and the sampling interval information is accumulated to form the anomaly holding duration information. When the illumination anomaly determination result information changes from abnormal to non-abnormal, the current segment ends, the anomaly holding duration information is cleared, and the accumulation processing of the next segment begins. When the anomaly holding duration information reaches the preset duration information, illumination anomaly warning information is generated, and the sampling time information associated with the illumination anomaly warning information is recorded.

[0048] S70: Sends abnormal light warning information to the terminal, receives parameter configuration instructions returned by the terminal, and updates the monitoring parameters of the acquisition terminal based on the parameter configuration instructions.

[0049] In this embodiment, the terminal refers to a mobile terminal used to receive light anomaly warning information and send parameter configuration instruction information to the acquisition end. The parameter configuration instruction information refers to instruction data carrying the content of monitoring parameter update. The monitoring parameters refer to the set of parameters related to the acquisition end's execution of light intensity data acquisition and anomaly warning.

[0050] Specifically, after generating an abnormal lighting warning, the acquisition end converts the warning information into transmittable warning data and writes it into a transmission buffer. It establishes a wireless connection with the terminal and sends the warning data in the transmission buffer while the wireless connection remains valid. After receiving the warning data, the terminal parses the abnormal lighting warning information and generates parameter configuration instruction information. The parameter configuration instruction information includes monitoring parameter update value information and parameter identifier information used to characterize the monitoring parameter type. After receiving the parameter configuration instruction information, the acquisition end performs parsing processing to extract the parameter identifier information and the monitoring parameter update value information. Based on the parameter identifier information, it locates the monitoring parameter record item that matches the parameter identifier information in the local parameter table and writes the monitoring parameter update value information into the monitoring parameter record item to complete the monitoring parameter update. When acquiring light intensity data and attitude data in the future, the acquisition end performs sampling time scheduling and abnormal warning-related processing according to the updated monitoring parameters.

[0051] In one embodiment, in step S20, the illumination intensity data and pose data are input into a trained data validity discrimination model to obtain validity labeling information, including: S201: Perform change rate calculation processing on the illumination intensity data within a preset time window to obtain illumination change characteristic information, and perform stability calculation processing on the attitude data within a preset time window to obtain attitude stability characteristic information.

[0052] In this embodiment, the illumination change feature information refers to the feature result used to characterize the rate of change of illumination intensity over time within a preset time window, and the attitude stability feature information refers to the feature result used to characterize the magnitude of change of attitude data over time within a preset time window.

[0053] Specifically, the start time of the preset time window is determined to be ts and the end time is te, and the conditions that satisfy the preset time window information are extracted. Sampling time sequence And the light intensity sample value sequence corresponding one-to-one with the sampling time sequence and attitude sampled value sequence The attitude sample value sequence is written as: when attitude data is represented in attitude angles and Represents pitch angle sampled value information and Represents the roll angle sample value information and This represents the yaw angle sampled value information. The sequence of attitude sampled values ​​is written as follows when the attitude data is represented as an attitude vector: and For three-dimensional component sampled value information, calculate the adjacent sampling interval sequence. and And in Time to discard and Associated adjacent pairs to ensure temporal order validity, calculate the adjacent illumination difference sequence. and And calculate the instantaneous rate of change of illumination sequence. and And calculate the absolute rate of change series The characteristic information of illumination change within a preset time window is calculated as a set of statistics, including the average absolute rate of change. With the maximum absolute rate of change ,in and When it is necessary to decouple the characteristic information of illumination change from the duration difference of different preset time windows, the window duration T=te-ts is calculated, and the change per unit time is calculated. As a supplementary component to the illumination change feature information, the difference between adjacent attitudes is then calculated from the attitude sampled value sequence to form an attitude change amplitude sequence. This calculation is performed when the attitude sampled value sequence is represented by attitude angles. , and And unify the angle difference to The equivalent range to avoid crossing The jump interference and angle normalization rule are as follows Calculate the attitude change amplitude And calculate the attitude change amplitude without using yaw angle sampled value information. When the attitude sample value sequence is represented by an attitude vector, the component difference is calculated. and And calculate the attitude change amplitude Based on attitude change amplitude sequence The attitude stability feature information is calculated as a set of statistics including the average change magnitude. With the maximum range of change With variance ,in and and When it is necessary to reflect the impact of time interval differences on stability, the time-weighted average change magnitude is calculated. Sw is used as a supplementary component of the pose stabilization feature information, and finally... As characteristic information of light change and Output as attitude stability feature information.

[0054] S202: Input the illumination change feature information and attitude stability feature information into the trained data validity discrimination model to obtain the validity label information corresponding to the preset time window.

[0055] Specifically, the process involves reading preset time window information and determining the corresponding window identifier, start time, and end time. It also reads the illumination change feature information and attitude stability feature information corresponding to the preset time window. Following the preset input field order of the trained data validity discrimination model, the illumination change feature information and attitude stability feature information are concatenated to form a discrimination feature vector. The input field order is the feature arrangement order fixed during the training phase of the trained data validity discrimination model. Based on the input field order, a dimensionality consistency check is performed on the discrimination feature vector information to confirm that the number of features in the discrimination feature vector information is consistent with the input dimension of the trained data validity discrimination model. If the validity verification process passes, the discriminant feature vector information is written into the model input cache, and the trained data validity discrimination model is triggered to perform inference calculation. The inference calculation includes performing a step-by-step comparison of the discriminant feature vector information according to the discrimination rules of the trained data validity discrimination model, performing path selection along the discrimination path corresponding to the comparison result until the leaf node is reached, and outputting the label value information associated with the leaf node. The label value information is determined as the validity label value information corresponding to the preset time window, and the window identifier information, window start time information, and window end time information are bound and recorded with the validity label value information. The validity label value information corresponding to all windows is aggregated along the window order of the preset time window information to form the validity label information corresponding to the preset time window.

[0056] In one embodiment, step S30, namely, filtering out invalid data from the illumination intensity data based on the validity labeling information to obtain valid illumination data, includes: S301: Traverse the light intensity data according to the sampling time order and read the validity marker information corresponding to the sampling time.

[0057] Specifically, the sampling time information and light intensity sample value information in the light intensity data are read. A time sequence verification process is performed on the sampling time information to confirm that it satisfies the monotonically increasing condition. If the time sequence verification fails, the sampling time information is reordered, and the light intensity sample value information associated with the sampling time information is adjusted synchronously according to the reordering result to maintain a one-to-one correspondence. Based on the verified or reordered sampling time information, a traversal index sequence is constructed, and the sampling time information corresponding to the current index is read sequentially along the order of the traversal index sequence. The sampling time information is used as the query key information to perform a matching retrieval process in the validity marker information. The matching retrieval process includes performing an interval determination on the window start time information and window end time information recorded in the validity marker information to determine the preset time window information into which the sampling time information falls, and reading the validity marker value information bound to the preset time window information as the validity marker information corresponding to the sampling time. When the validity marker information adopts a sampling time-by-sampling method, the matching retrieval process includes directly reading the validity marker value information in the validity marker information with the sampling time information as an index as the validity marker information corresponding to the sampling time, and writing the sampling time information, light intensity sampling value information and validity marker information into the traversal record sequence information for subsequent screening processing.

[0058] S302: If the validity marker information indicates that the light intensity data corresponding to the sampling time is determined as a valid sample value and retained.

[0059] Specifically, the sampling time information is read from the traversal record sequence information, and the light intensity sampling value information and validity mark information corresponding one-to-one with the sampling time information are read. The validity mark information is processed to confirm that the validity mark information indicates validity. If the validity mark information indicates validity, the light intensity sampling value information is assigned as valid sampling value information, and the sampling time information and valid sampling value information are bound and written to the valid sampling value cache. When the valid sampling value cache reaches the preset cache length, the sampling time information and valid sampling value information in the valid sampling value cache are appended and written to the valid light illumination data storage area in the order of the sampling time information. When the valid sampling value cache does not reach the preset cache length, the valid sampling value cache is maintained to continue receiving valid sampling value information corresponding to subsequent sampling time information. After the traversal is completed, the remaining sampling time information and valid sampling value information in the valid sampling value cache are appended and written to the valid light illumination data storage area in the order of the sampling time information to complete the retention process of valid sampling value information.

[0060] S303: If the validity marker information indicates that the data is invalid, the light intensity data corresponding to the sampling time shall be identified as invalid data and filtered out. The remaining valid sampled values ​​shall be collected in the order of sampling time to form valid light data.

[0061] Specifically, the sampling time information is read from the traversal record sequence information, and the light intensity sample value information and validity marker information corresponding one-to-one with the sampling time information are read. The validity marker information is processed to determine whether the validity marker information indicates invalidity. If the validity marker information indicates invalidity, the light intensity sample value information is assigned as invalid data and a filtering process is performed. The filtering process includes writing the sampling time information and invalid data into the invalid data recording area and writing a filtering marker value into the record items in the invalid data recording area to indicate that invalid data will not participate in subsequent feature extraction processing, while maintaining the existing valid sample values ​​in the valid sample value cache. Information is not deleted to avoid disrupting the sampling time order. After traversal, all valid sampled values ​​in the valid illumination data storage area are read and sorted according to the sampling time information to confirm that the sampling time information meets the monotonically increasing condition. If the sorting verification fails, the sampling time information is re-sorted to form a valid sampled value sequence arranged in the sampling time order. The valid sampled value sequence arranged in the sampling time order is written into the valid illumination data result area to form valid illumination data. If the sorting verification passes, the valid sampled value information in the valid illumination data storage area is directly confirmed as valid illumination data.

[0062] In one embodiment, step S40, which involves performing feature extraction processing on the effective illumination data to obtain illumination feature information, includes: S401: Perform segmented processing on the effective illumination data according to the preset time window information to obtain the time window sequence information.

[0063] Specifically, the process involves reading preset time window information and determining the preset time window length and window sliding step size. It also involves reading sampling time information and valid sample value information from the effective illumination data and performing time sequence verification on the sampling time information to confirm that the sampling time information satisfies the monotonically increasing condition. Based on the sampling time information, the start time information t0 and end time information tM of the effective illumination data are determined. The first preset time window information is constructed using the start time information t0 as the start time information of the first window and t0+W as the end time information of the first window, where W represents the preset time window length. Subsequently, the window start time information is recursively updated according to the window sliding step size S to obtain the start time information of the kth window. And obtain the information of the termination time tk+W of the k-th window. Under the condition of tk < tM, loop to generate the preset time window information until the information of the window start time is not less than the information of the termination time. For each preset time window information, perform interval inclusion determination to filter the valid sampling value information corresponding to the sampling time information that meets and form the sequence information of valid sampling values within the window, and perform binding record processing on the window identification information, the window start time information, the window termination time information, and the sequence information of valid sampling values within the window to form the time window entry information. Finally, perform aggregation processing on all the time window entry information in the order of the window start time information to obtain the time window sequence information S402: In the time window sequence information, perform cumulative statistical processing on the valid light data within each preset time window information to obtain the information of the valid light duration.

[0064] In this embodiment, the information of the valid light duration refers to the cumulative continuous duration in the time dimension of the valid sampling values represented by the valid light data within the time range corresponding to the preset time window information.

[0065] Specifically, read the time window sequence information and sequentially read the window start time information ts, the window termination time information te, and the sequence information of valid sampling values within the window corresponding to each preset time window information. The sequence information of valid sampling values within the window includes a sampling time sequence arranged in the order of sampling times and a sequence of valid sampling values corresponding one by one to the sampling time sequence , when N = 0, record the information of the valid light duration as 0, when N = 1, record the information of the valid light duration as non-negative result of At successively calculate the adjacent sampling intervals and perform non-negative check on discard the corresponding adjacent pairs to maintain the validity of the sampling time order, perform interval clipping processing on the remaining adjacent pairs to obtain the valid intervals within the window , accumulate all the valid intervals within the window to obtain the window valid light duration and bind the window valid light duration Tw to the window identification information to form the information of the valid light duration.

[0066] S403: Perform mean calculation processing on the valid light data within the preset time window information to obtain the average light intensity information.

[0067] In this embodiment, the average light intensity information refers to the mean result of weighted valid sampling values in the time dimension within the time range corresponding to the preset time window information.

[0068] Specifically, the preset time window information is read and the start time information of the window is determined as ts and the end time information of the window is te. The valid illumination data within the preset time window information is read and the sampling time sequence arranged in the order of sampling time is extracted. A sequence of valid sampled values ​​corresponding one-to-one with the sampling time sequence When N=0, the average light intensity information is recorded as 0. Calculate the adjacent sampling interval at time and And on Perform nonnegativity check to Discarding adjacent pairs at the same time to maintain the validity of the sampling time order, and performing window pruning on the retained adjacent pairs to obtain the effective interval within the window. Calculate the effective interval within the end window based on the end sampling time information. The effective sampled value sequence is bound to the effective intervals within the window according to the segmented constant value rule to construct a time-weighted entry set and calculate the interval cumulative value information. Calculate the weighted cumulative value information Calculate the average illuminance information when D>0. and output As the average illuminance information, the average illuminance information is recorded as 0 when D=0. S404: Perform threshold comparison statistical processing on the effective illumination data within the preset time window to obtain strong light exposure frequency information and weak light exposure frequency information. The strong light exposure frequency information is the statistical result of the number of times the effective illumination data is greater than the second threshold information, and the weak light exposure frequency information is the statistical result of the number of times the effective illumination data is less than the first threshold information.

[0069] Specifically, the system reads the preset time window information and determines the window start time information ts and the window end time information te. It also reads the preset light health threshold range and parses it to obtain the first threshold information T1 and the second threshold information T2. ​​Within the time range corresponding to the preset time window information, it reads valid light data and extracts the sampling time sequence. With valid sample value sequence When N=0, the frequency information of strong light exposure is recorded as 0, and the frequency information of weak light exposure is also recorded as 0. For each valid sample value Li, a threshold comparison process is performed to generate a strong light comparison marker value. Compare the marked values ​​with those in low light. The strong light comparison mark value satisfies If and only if Li > T2 and in Time to take , the low - light comparison marker value satisfies if and only if Li < T1 and at take , perform cumulative statistics on the high - light comparison marker value to obtain high - light exposure frequency information , perform cumulative statistics on the low - light comparison marker value to obtain low - light exposure frequency information , bind the high - light exposure frequency information FH and the low - light exposure frequency information FL to the preset time - window information to form the statistical result corresponding to the preset time - window information S405: Collect the effective light duration information, average light intensity information, high - light exposure frequency information, and low - light exposure frequency information to form light - feature information.

[0070] Specifically, read the time - window sequence information and determine the window identification information, window start - time information, and window end - time information corresponding to each preset time - window information. For the same window identification information, read the effective light duration information, average light intensity information, high - light exposure frequency information, and low - light exposure frequency information respectively, and perform integrity verification processing on the effective light duration information, average light intensity information, high - light exposure frequency information, and low - light exposure frequency information to confirm that all four types of information exist and correspond one - to - one with the window identification information. When the integrity verification processing passes, perform field splicing processing on the four types of information according to the preset field - order information to form light - feature entry information. The preset field - order information is a field arrangement order where the effective light duration information field is in the front, the average light intensity information field is behind, and the high - light exposure frequency information field and the low - light exposure frequency information field are at the end. Perform binding record processing on the window identification information, window start - time information, window end - time information, and light - feature entry information to form window - level light - feature record information. Perform collection processing on all window - level light - feature record information in the time order of the window start - time information to form light - feature sequence information, and determine the light - feature sequence information as light - feature information.

[0071] In one embodiment, in step S403, that is, perform mean - value calculation processing on the effective light data within the preset time - window information to obtain the average light intensity information, including: S4031: Obtain the effective light data arranged in the order of sampling moments within the preset time - window information, and extract the sampling - moment information and light - intensity sampling - value information from the effective light data.

[0072] Specifically, the preset time window information is read, and the window start time information ts and window end time information te corresponding to the preset time window information are determined. Valid illumination data is read, and the set of valid sampled values ​​that correspond one-to-one with the set of sampled time information recorded in the valid illumination data are extracted. Based on the window start time information ts and window end time information te, interval filtering processing is performed on the set of sampled time information to obtain the desired result. Sampling time sequence within the window According to the sampling time information sequence within the window In the valid illumination data, indexing and positioning processing is performed to read the illumination intensity sample value information Li corresponding one-to-one with each ti, and to form a sequence of illumination intensity sample values ​​within the window. The sequence of sampling time information within the window Perform time sequence verification and, if the time sequence verification fails, sample the time sequence information within the window. Perform ascending reordering and sample the sequence of light intensity values ​​within the window. The index is synchronously reordered to maintain a one-to-one correspondence, and the sampling time sequence within the window is reordered. The sequence of light intensity samples within the window was confirmed as the sampling time information and then reordered. The light intensity sampled value information is confirmed, and the sampling time information and the light intensity sampled value information are combined in the order of sampling time to form the effective light data arranged in the order of sampling time within the preset time window information.

[0073] S4032: Calculate the sampling interval information based on the information of adjacent sampling times, and bind the sampling interval information with the light intensity sampling value information corresponding to the information of adjacent sampling times to obtain the time-weighted entry information.

[0074] In this embodiment, time-weighted entry information refers to the set of entries formed by using the sampling interval information determined by adjacent sampling time information as the time weight and establishing a correspondence with the light intensity sampling value information corresponding to the adjacent sampling time information.

[0075] Specifically, the sampling time information is read and the sampling time sequence arranged in order of sampling time is recorded as follows: Read the light intensity sample value information and record the light intensity sample value sequence corresponding one-to-one with the sampling time sequence as When N < 2, the time-weighted entry information is set to an empty set. Next-order sampling time pairing processing is performed from index i=1 to i=N-1 to obtain adjacent sampling time information. The sampling interval information is calculated based on the information of adjacent sampling times and recorded as . ,right Perform non-negative check processing and Time to discard and Associated adjacent sampling pairs to ensure the validity of the sampling time sequence, in Sampling interval information The light intensity sampled values ​​corresponding to adjacent sampling times are bound together and formed into entries. The light intensity sampling value information is selected from the light intensity sampling value Li corresponding to the starting sampling time ti in the adjacent sampling time information to characterize the light intensity sampling value. The light intensity values ​​within the time range are collected, and finally all entries are aggregated in ascending order of ti to form time-weighted entry information.

[0076] S4033: Perform illumination change rate calculation on the time-weighted entry information to obtain the corrected sampling interval information.

[0077] In this embodiment, the corrected sampling interval information refers to the interval correction result formed by combining the sampling interval information with the calculation result of the illumination change rate, which is used to characterize the effective time weight related to the illumination change amplitude.

[0078] Specifically, the time-weighted entry information is read and adjacent sampling time information is extracted in order of entry. Sampling interval information And the illumination intensity sample value information Li corresponding to ti, and extract the information from the time-weighted entry information. Corresponding light intensity sampling information Illumination difference information is calculated based on the light intensity sampled value information. Illumination change rate information is calculated based on illumination difference information and sampling interval information. Interval correction factor information is constructed based on illumination change rate information. The sampling interval information is corrected by using the interval correction factor information to obtain the corrected sampling interval information. The time-weighted entry information is arranged in the order of all entries. Perform the aggregation and establish a one-to-one correspondence with ti in the original entries to form the corrected sampling interval information. S4034: Perform time-weighted accumulation processing on the light intensity sampled value information based on the corrected sampling interval information to obtain the weighted accumulation value information, and perform accumulation processing on the corrected sampling interval information to obtain the interval accumulation value information.

[0079] In this embodiment, the weighted cumulative value information refers to the cumulative result formed by adding the light intensity sampled value information with time weight according to the corrected sampling interval information, and the interval cumulative value information refers to the cumulative result formed by adding the corrected sampling interval information with time weight.

[0080] Specifically, the time-weighted entry information is read and the sequence of light intensity sampled values ​​corresponding one-to-one with the entry index is extracted. and the corrected sampling interval information sequence ,in Information from adjacent sampling times A one-to-one correspondence is established between the light intensity sampled value information L_i corresponding to t_i. When N<2, the weighted cumulative value information is set to 0 and the interval cumulative value information is set to 0. Time Perform non-negative check processing and When To maintain numerical stability during the accumulation process, the weighted accumulation value information is initialized to S=0 and the interval accumulation value information is initialized to D=0. The item-by-item accumulation process is performed from i=1 to i=N-1, and the item weighting term is calculated in the i-th item. And perform weighted cumulative value update. Simultaneously, perform interval accumulation value update. After the accumulation is completed, S is determined as the weighted accumulated value information and D is determined as the interval accumulated value information. S4035: Perform ratio calculation processing using the weighted cumulative value information and the interval cumulative value information to obtain the average light intensity information.

[0081] Specifically, the weighted cumulative value information is read and recorded as S, and the interval cumulative value information is read and recorded as D. Zero-value determination processing is performed on the interval cumulative value information D to confirm whether D is zero. When D=0, the average illuminance information is recorded as 0. When D>0, ratio calculation processing is performed to calculate the average illuminance information. The ratio calculation process includes performing numerical type consistency processing on S and D to convert them into floating-point values ​​of the same precision, performing division to obtain the ratio result information, and performing significant digit truncation processing on the ratio result information to obtain the output format value information of the average illuminance information. Finally, the ratio is... The average light intensity information is determined and bound to the preset time window information for recording.

[0082] In one embodiment, in step S50, anomaly detection processing is performed based on illumination feature information and a preset illumination health threshold range to obtain illumination anomaly detection result information, including: S501: Read the preset light health threshold range and parse the preset light health threshold range into first threshold information and second threshold information.

[0083] Specifically, the system accesses the monitoring parameter storage area associated with the acquisition terminal and reads the threshold interval record information corresponding to the preset light health threshold interval. The threshold interval record information includes the lower limit field information, the upper limit field information, and the field unit information. The lower limit value is extracted based on the lower limit field information and determined as the first threshold information. The upper limit value is extracted based on the upper limit field information and determined as the second threshold information. The first threshold information and the second threshold information are subjected to numerical validity verification to confirm that the first threshold information is less than the second threshold information and that the first threshold information and the second threshold information meet the preset value range. If the numerical validity verification fails, the first threshold information and the second threshold information are reset using the default value field information in the threshold interval record information. If the numerical validity verification passes, the first threshold information and the second threshold information are written into the threshold cache and a binding record is established with the preset light health threshold interval to complete the parsing process.

[0084] S502: Extract average light intensity information, strong light exposure frequency information, and weak light exposure frequency information from the light characteristic information.

[0085] Specifically, the process involves reading illumination feature information and determining the corresponding field mapping relationships. These field mapping relationships characterize the correspondence between the positions and names of each field in the illumination feature information. Based on these relationships, the process locates the field corresponding to the average illumination intensity information within the illumination feature information and reads its value to extract the average illumination intensity information. Similarly, it locates the field corresponding to the strong light exposure frequency information within the illumination feature information and reads its value to extract the strong light exposure frequency information. Finally, it locates the field corresponding to the weak light exposure information within the illumination feature information. The system locates the corresponding fields for frequency information and reads their values ​​to extract low-light exposure frequency information. It then performs data type validation on the extracted average light intensity, strong light exposure frequency, and low-light exposure frequency information to confirm that the average light intensity is a numeric type and that the strong and low-light exposure frequency information are non-negative integers. If the data type validation fails, it performs format conversion on the corresponding field values ​​and re-validates them. Finally, it writes the average light intensity, strong light exposure frequency, and low-light exposure frequency information into the judgment input buffer for subsequent comparison processing.

[0086] S503: Perform a comparison process between the average light intensity information and the first threshold information to generate weak light anomaly marker information when the average light intensity information is less than the first threshold information.

[0087] Specifically, the average illuminance information and the first threshold information are read. A unit consistency check is performed on the average illuminance information and the first threshold information to confirm that they use the same illuminance unit information. If the unit consistency check fails, the average illuminance information is converted according to the unit conversion rule information to obtain the converted average illuminance information. Then, the converted average illuminance information and the first threshold information are compared numerically. If the converted average illuminance information is less than the first threshold information, the weak light anomaly marker information is written as an abnormal value and the abnormal value is bound to the trigger time information. If the converted average illuminance information is not less than the first threshold information, the weak light anomaly marker information is written as a non-abnormal value and the non-abnormal value is bound to the trigger time information. The trigger time information is taken from the window end time information or window start time information of the preset time window information corresponding to the average illuminance information.

[0088] S504: Perform a comparison process between the average light intensity information and the second threshold information to generate strong light anomaly marker information when the average light intensity information is greater than the second threshold information.

[0089] Specifically, the average illuminance information and the second threshold information are read. A unit consistency check is performed on the average illuminance information and the second threshold information to confirm that they use the same illuminance unit information. If the unit consistency check fails, the average illuminance information is converted according to the unit conversion rule information to obtain the converted average illuminance information. Then, the converted average illuminance information and the second threshold information are compared numerically. If the converted average illuminance information is greater than the second threshold information, the strong light anomaly marker information is written as an abnormal value and the abnormal value is bound to the trigger time information. If the converted average illuminance information is not greater than the second threshold information, the strong light anomaly marker information is written as a non-abnormal value and the non-abnormal value is bound to the trigger time information. The trigger time information is taken from the window end time information or window start time information of the preset time window information corresponding to the average illuminance information.

[0090] S505: When at least one of the strong light anomaly marker information and the weak light anomaly marker information exists, perform an anomaly consistency verification process on the strong light exposure frequency information and the weak light exposure frequency information, so as to generate light anomaly judgment result information when the strong light exposure frequency information or the weak light exposure frequency information meets the preset exposure conditions.

[0091] Specifically, the system reads strong light anomaly marker information and weak light anomaly marker information and performs anomaly existence determination processing to confirm that at least one of the strong light anomaly marker information and weak light anomaly marker information is an anomalous value. If the anomaly existence determination processing passes, the system reads strong light exposure frequency information and weak light exposure frequency information, and reads preset exposure condition information. The preset exposure condition information includes strong light exposure frequency threshold information, weak light exposure frequency threshold information, and exposure determination mode information. The exposure determination mode information is used to characterize the determination method of using strong light exposure frequency information alone, weak light exposure frequency information alone, or a combination of strong light exposure frequency information and weak light exposure frequency information for determination. Based on the exposure determination mode information, the system performs comparison processing on the strong light exposure frequency information and the strong light exposure frequency threshold information to obtain strong light exposure satisfaction marker information, and performs comparison processing on the weak light exposure frequency information and the weak light exposure frequency threshold information to obtain weak light exposure satisfaction marker information. The system performs consistency verification processing based on strong light anomaly marker information and strong light exposure satisfaction marker information. When the strong light anomaly marker information is an anomalous value and the strong light exposure satisfaction marker information is a satisfactory value, it outputs strong light consistency pass marker information. It also performs consistency verification processing based on weak light anomaly marker information and weak light exposure satisfaction marker information. When the weak light anomaly marker information is an anomalous value and the weak light exposure satisfaction marker information is a satisfactory value, it outputs weak light consistency pass marker information. When at least one of the strong light consistency pass marker information and the weak light consistency pass marker information is a pass value, it generates illumination anomaly judgment result information and writes it into the anomaly type field information. The anomaly type field information is used to characterize strong light anomaly, weak light anomaly, or both strong light anomaly and weak light anomaly. When neither the strong light consistency pass marker information nor the weak light consistency pass marker information is a pass value, it writes the illumination anomaly judgment result information as a non-abnormal value and records the trigger time information associated with the non-abnormal value.

[0092] In one embodiment, in step S60, i.e., when the lighting anomaly determination result information indicates an anomaly, continuous determination processing is performed on the lighting anomaly determination result information to generate lighting anomaly warning information, including: S601: When the light abnormality determination result indicates an abnormality, obtain the preset duration information.

[0093] Specifically, the system reads the illumination anomaly determination result information and performs anomaly state determination processing to confirm that the illumination anomaly determination result information is an abnormal value. If the anomaly state determination processing passes, it accesses the monitoring parameter storage area associated with the acquisition end and locates the parameter record item corresponding to the preset duration information. The parameter record item includes duration value field information, duration unit field information, and parameter effective range field information. The system extracts the duration value based on the duration value field information and parses the duration unit based on the duration unit field information to obtain the preset duration information. Based on the parameter effective range field information, it performs applicability verification processing to confirm that the preset duration information is applicable to the anomaly type field information corresponding to the current illumination anomaly determination result information. If the applicability verification processing fails, it reads the backup duration record item that matches the anomaly type field information to reset the preset duration information. If the applicability verification processing passes, it writes the preset duration information into the continuity determination cache and binds it to the current illumination anomaly determination result information.

[0094] S602: Determine the continuity determination window information according to the preset duration information.

[0095] In this embodiment, the continuity determination window information refers to the time range definition information used to carry out the continuity determination process. The continuity determination window information is determined by the preset duration information and is time-aligned with the trigger time information of the illumination anomaly determination result information.

[0096] Specifically, the system reads the preset duration information and the trigger time information bound to the illumination anomaly judgment result information. The trigger time information is taken from the window end time information or window start time information of the preset time window information corresponding to the illumination anomaly judgment result information. The system determines the window end time information of the continuity judgment window information as the trigger time information and determines the window start time information of the continuity judgment window information as the trigger time information. It then backtracks the duration result corresponding to the preset duration information. Based on the sampling configuration parameter record information, the system reads the sampling period information and performs sampling period alignment processing on the preset duration information to obtain the aligned duration information. The alignment processing includes converting the preset duration information into an integer multiple of the sampling period information and rounding up when the conversion result has a decimal part to avoid missing continuous anomaly segments. Based on the aligned duration information, the system updates the window start time information of the continuity judgment window information to complete the determination of the continuity judgment window information. Finally, the system writes the continuity judgment window information into the continuity judgment cache so that a consistent window boundary can be used when reading the illumination anomaly judgment result information in the future.

[0097] S603: Within the sampling time range corresponding to the continuity determination window information, read the illumination anomaly determination result information sequentially along the sampling time.

[0098] Specifically, the continuity judgment window information is read and the window start time and window end time information are extracted. The historical record area of ​​the illumination anomaly judgment result information is accessed and the set of record items falling between the window start time information and window end time information is located. The set of record items contains sampling time information, judgment value field information, and anomaly type field information. The time order verification process is performed on the set of record items according to the sampling time information to confirm that the sampling time information meets the monotonically increasing condition. If the time order verification process fails, the set of record items is reordered in ascending order according to the sampling time information, and the one-to-one correspondence between each sampling time information and the corresponding judgment value field information and anomaly type field information is maintained. Each record item is read sequentially along the reordered sampling time information and the judgment value field information is determined as the illumination anomaly judgment result information corresponding to the sampling time. At the same time, the sampling time information and the illumination anomaly judgment result information are written into the continuity judgment sequence cache to form the in-window judgment sequence information. The in-window judgment sequence information is used for subsequent anomaly retention time statistics processing.

[0099] S604: Statistically calculate the duration of abnormality within the sampling time range corresponding to the continuity determination window information.

[0100] In this embodiment, the abnormality retention duration information refers to the cumulative duration during which the illumination abnormality determination result information retains the abnormal value within the sampling time range corresponding to the continuity determination window information.

[0101] Specifically, the continuity determination window information is read and the window start time and window end time information are extracted. The determination sequence information within the window is read and the sequence of sampling time information arranged in the sampling time order is extracted, along with the corresponding sequence of illumination anomaly determination results. The anomaly retention duration information is initialized to zero and the anomaly segment start time information is initialized to null. Illumination anomaly determination result information is read sequentially along the sampling time order, and anomaly value determination processing is performed. When the anomaly value determination processing indicates an anomaly and the anomaly segment start time information is null, the current sampling time information is written into the anomaly segment start time information to indicate the start of the anomaly segment. When the anomaly value determination processing indicates an anomaly and the anomaly segment start time information is not null, the anomaly segment start time information is kept unchanged to indicate the continuation of the anomaly segment. When the anomaly value determination processing indicates a non-anomaly and the anomaly segment starts... When the initial time information is not null, the current sampling time information is determined as the termination time information of the abnormal segment, and the segment duration calculation is performed to obtain the segment holding duration information. The segment duration calculation uses the time difference between the termination time information of the abnormal segment and the start time information of the abnormal segment as the segment holding duration information, and performs non-negative verification on the time difference to remove abnormal sequence records. Then, the segment holding duration information is added to the abnormal holding duration information, and the abnormal segment start time information is cleared to end the abnormal segment statistics. When the traversal ends and the abnormal segment start time information is not null, the window termination time information is determined as the termination time information of the abnormal segment, and the segment duration calculation is performed. The obtained segment holding duration information is added to the abnormal holding duration information. Finally, the abnormal holding duration information and the continuity determination window information are bound and recorded for subsequent duration determination.

[0102] S605: When the duration of abnormality reaches the preset duration, generate an abnormal lighting warning.

[0103] In this embodiment, the light anomaly warning information refers to the warning trigger record information formed after the continuous determination process meets the duration condition. The light anomaly warning information includes at least the warning trigger time information, the anomaly type field information, and the warning level field information.

[0104] Specifically, the system reads the abnormality persistence duration information and the preset duration information. It then performs a unit consistency check on the abnormality persistence duration information and the preset duration information to confirm that they use the same time unit. If the unit consistency check fails, it performs a unit conversion process on the abnormality persistence duration information according to the time unit conversion rules to obtain the converted abnormality persistence duration information. Subsequently, it performs a threshold determination process on the converted abnormality persistence duration information and the preset duration information. When the converted abnormality persistence duration information reaches the preset duration information, it reads the abnormality type field information of the illumination abnormality determination result information corresponding to the current continuity determination window information and reads the preset warning configuration parameters information. The preset warning configuration parameters information includes warning level mapping rule information and warning suppression interval information. The system includes information on warning trigger frequency limits, and performs suppression judgment processing based on the warning suppression interval information and the warning trigger time information of the most recent light anomaly warning information in the historical warning record area to avoid repeatedly generating light anomaly warning information within the suppression interval. If the suppression judgment processing passes, the anomaly type field information is mapped to the warning level field information according to the warning level mapping rule information, and the window termination time information of the continuity judgment window information is determined as the warning trigger time information. The warning trigger time information, anomaly type field information, warning level field information and associated window identifier information are packaged together to form light anomaly warning information and written into the historical warning record area. If the suppression judgment processing fails, the light anomaly warning information is written as a non-trigger value and the suppression reason field information is recorded.

[0105] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0106] In one embodiment, a machine learning-based illumination feedback device is provided, which corresponds one-to-one with the machine learning-based illumination feedback method described in the above embodiments. For example... Figure 2 As shown, this machine learning-based illumination feedback device includes a data acquisition module, a validity determination module, an invalid filtering module, a feature extraction module, an anomaly detection module, a continuous early warning module, and a parameter update module. Detailed descriptions of each functional module are as follows: In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Acquire the light intensity data collected by the acquisition terminal, and acquire the attitude data corresponding to the time of the light intensity data; The illumination intensity data and pose data are input into the trained data validity discrimination model to obtain validity labeling information; Invalid data is filtered out from the light intensity data based on the validity labeling information to obtain valid light intensity data; Feature extraction processing is performed on the effective illumination data to obtain illumination feature information; Anomaly detection processing is performed based on illumination feature information and a preset illumination health threshold range to obtain illumination anomaly detection result information; When the lighting anomaly determination result indicates an anomaly, continuous determination processing is performed on the lighting anomaly determination result to generate lighting anomaly early warning information; The system sends abnormal light warning information to the terminal, receives parameter configuration instructions from the terminal, and updates the monitoring parameters of the acquisition terminal based on the parameter configuration instructions.

[0107] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Acquire the light intensity data collected by the acquisition terminal, and acquire the attitude data corresponding to the time of the light intensity data; The illumination intensity data and pose data are input into the trained data validity discrimination model to obtain validity labeling information; Invalid data is filtered out from the light intensity data based on the validity labeling information to obtain valid light intensity data; Feature extraction processing is performed on the effective illumination data to obtain illumination feature information; Anomaly detection processing is performed based on illumination feature information and a preset illumination health threshold range to obtain illumination anomaly detection result information; When the lighting anomaly determination result indicates an anomaly, continuous determination processing is performed on the lighting anomaly determination result to generate lighting anomaly early warning information; The system sends abnormal light warning information to the terminal, receives parameter configuration instructions from the terminal, and updates the monitoring parameters of the acquisition terminal based on the parameter configuration instructions.

[0108] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0109] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A machine learning-based illumination feedback method, characterized in that, The machine learning-based illumination feedback method includes: Acquire light intensity data collected by the acquisition terminal, and acquire attitude data corresponding to the time of the light intensity data; The illumination intensity data and the attitude data are input into a trained data validity discrimination model to obtain validity labeling information; Based on the validity labeling information, invalid data is filtered out from the light intensity data to obtain valid light data; Perform feature extraction processing on the effective illumination data to obtain illumination feature information; Based on the light feature information and the preset light health threshold range, anomaly detection processing is performed to obtain light anomaly detection result information; If the lighting anomaly determination result indicates an anomaly, continuous determination processing is performed on the lighting anomaly determination result to generate lighting anomaly warning information; The abnormal light warning information is sent to the terminal, the parameter configuration instruction information returned by the terminal is received, and the monitoring parameters of the acquisition terminal are updated based on the parameter configuration instruction information.

2. The illumination feedback method based on machine learning according to claim 1, characterized in that, The step of inputting the illumination intensity data and the pose data into a trained data validity discrimination model to obtain validity labeling information includes: The light intensity data is processed by performing a rate of change calculation within a preset time window to obtain light change characteristic information, and the attitude data is processed by performing a stability calculation within the preset time window to obtain attitude stability characteristic information. The illumination change feature information and the attitude stability feature information are input into the trained data validity discrimination model to obtain validity label information corresponding to the preset time window.

3. The illumination feedback method based on machine learning according to claim 1, characterized in that, The process of filtering out invalid data from the illumination intensity data based on the validity labeling information to obtain valid illumination data includes: The light intensity data is traversed in the order of sampling time, and the validity marker information corresponding to the sampling time is read. If the validity marker information indicates that the sample is valid, the light intensity data corresponding to the sampling time will be determined as a valid sample value and retained. If the validity marker information indicates invalidity, the light intensity data corresponding to the sampling time is determined as invalid data and filtered out. The remaining valid sampled values ​​are collected in the order of the sampling time to form the valid light data.

4. The illumination feedback method based on machine learning according to claim 1, characterized in that, The step of performing feature extraction processing on the effective illumination data to obtain illumination feature information includes: The effective illumination data is segmented according to the preset time window information to obtain the time window sequence information; In the time window sequence information, cumulative statistical processing is performed on the effective illumination data within each preset time window information to obtain effective illumination duration information; Within the preset time window, the effective illumination data is subjected to mean calculation to obtain average illumination intensity information; Within the preset time window, threshold comparison statistical processing is performed on the effective illumination data to obtain strong light exposure frequency information and weak light exposure frequency information. The strong light exposure frequency information is the statistical result of the number of times the effective illumination data is greater than the second threshold information, and the weak light exposure frequency information is the statistical result of the number of times the effective illumination data is less than the first threshold information. The effective illumination duration information, the average illumination intensity information, the strong light exposure frequency information, and the weak light exposure frequency information are combined to form the illumination characteristic information.

5. The illumination feedback method based on machine learning according to claim 4, characterized in that, The step of performing mean calculation processing on the effective illumination data within the preset time window to obtain average illumination intensity information includes: Obtain valid illumination data arranged in the order of sampling time within the preset time window, and extract the sampling time information and illumination intensity sampling value information from the valid illumination data; The sampling interval information is calculated based on the adjacent sampling time information, and the sampling interval information is bound with the light intensity sampling value information corresponding to the adjacent sampling time information to obtain the time-weighted entry information; The time-weighted entry information is processed by calculating the rate of change of illumination to obtain the corrected sampling interval information; Based on the corrected sampling interval information, time-weighted accumulation processing is performed on the light intensity sample value information to obtain weighted accumulation value information, and accumulation processing is performed on the corrected sampling interval information to obtain interval accumulation value information; The average light intensity information is obtained by performing a ratio calculation process using the weighted cumulative value information and the interval cumulative value information.

6. The illumination feedback method based on machine learning according to claim 1, characterized in that, The abnormality detection process is performed based on the illumination feature information and a preset illumination health threshold range to obtain illumination abnormality detection result information, including: Read the preset light health threshold range and parse the preset light health threshold range into first threshold information and second threshold information; The average light intensity information, strong light exposure frequency information, and weak light exposure frequency information are extracted from the light characteristic information. The average light intensity information is compared with the first threshold information to generate weak light anomaly marker information when the average light intensity information is less than the first threshold information. The average light intensity information is compared with the second threshold information to generate strong light anomaly marker information when the average light intensity information is greater than the second threshold information. When at least one of the strong light anomaly marker information and the weak light anomaly marker information exists, the strong light exposure frequency information and the weak light exposure frequency information are subjected to anomaly consistency verification processing, so as to generate the light anomaly judgment result information when the strong light exposure frequency information or the weak light exposure frequency information meets the preset exposure conditions.

7. The illumination feedback method based on machine learning according to claim 1, characterized in that, When the illumination anomaly determination result information indicates an anomaly, the step of performing continuous determination processing on the illumination anomaly determination result information to generate illumination anomaly early warning information includes: When the lighting anomaly determination result indicates an anomaly, obtain the preset duration information; The continuity determination window information is determined according to the preset duration information; Within the sampling time range corresponding to the continuity determination window information, the illumination anomaly determination result information is read sequentially along the sampling time. Within the sampling time range corresponding to the continuity determination window information, the duration of abnormality retention is statistically analyzed; When the duration of the abnormality reaches the preset duration, the light abnormality warning information is generated.

8. A light feedback device based on machine learning, characterized in that, The machine learning-based illumination feedback device includes: The data acquisition module is used to acquire light intensity data collected by the acquisition terminal and to acquire attitude data corresponding to the time of the light intensity data. The validity discrimination module is used to input the illumination intensity data and the posture data into the trained data validity discrimination model to obtain validity labeling information; An invalid data filtering module is used to filter out invalid data from the light intensity data based on the validity labeling information to obtain valid light intensity data; The feature extraction module is used to perform feature extraction processing on the effective illumination data to obtain illumination feature information; The anomaly detection module is used to perform anomaly detection processing based on the illumination feature information and the preset illumination health threshold range to obtain illumination anomaly detection result information; The continuous early warning module is used to perform continuous judgment processing on the light anomaly judgment result information to generate light anomaly early warning information when the light anomaly judgment result information indicates an anomaly. The parameter update module is used to send the abnormal light warning information to the terminal, receive the parameter configuration instruction information returned by the terminal, and update the monitoring parameters of the acquisition terminal based on the parameter configuration instruction information.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the machine learning-based illumination feedback method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the machine learning-based illumination feedback method as described in any one of claims 1 to 7.

Images