Abnormal behavior trigger type attendance recording video control system based on video linkage

Abstract

The present application relates to the technical field of video recording control, in particular to an abnormal behavior trigger type attendance video recording control system based on video linkage. It comprises: an event acquisition and integration unit, which acquires recognition failure events in real time and generates an exponentially decaying cumulative intensity value; an environment perception unit, which acquires environmental parameters such as illumination intensity; an adaptive threshold determination unit, which dynamically adjusts the trigger threshold according to the environmental parameters and outputs an abnormality level when the cumulative intensity value exceeds the threshold; a video recording instruction generation unit, which generates differentiated control instructions according to the abnormality level; a video recording execution unit, which stores the video recording of the abnormal behavior period in a designated storage area; and an abnormality auxiliary control unit, which synchronously executes evidence locking and light enhancement. The system can achieve sensitive scene anomaly detection, cache key pictures at low cost before abnormality triggering, and automatically adjust the video recording quality and storage strategy, thereby avoiding false positives and false negatives and improving evidence integrity and resource utilization.

Description

Technical Field

[0001] This invention relates to the field of video recording control technology, and more specifically, to an abnormal behavior-triggered attendance recording control system based on video linkage. Background Technology

[0002] Facial recognition attendance systems are widely used in enterprises, schools, and industrial parks. To assist in post-event traceability and evidence collection for abnormal behavior, video recording equipment is usually deployed at attendance points. Ideally, the recording system should initiate high-quality recording and save key evidence when abnormal behavior occurs, while maintaining low power consumption or only performing loop buffering during normal attendance. However, "abnormal behavior" in attendance scenarios often presents characteristics such as short-term suddenness, high susceptibility to environmental interference, and difficulty in distinguishing it from normal failures, posing challenges to real-time detection and recording control.

[0003] To reduce the storage pressure caused by continuous recording, existing technologies have developed recording triggering schemes based on motion detection or face detection. However, this method cannot distinguish between normal passage and malicious attempts, and frequent daily attendance will also trigger a large number of invalid recordings, resulting in limited storage savings, and subsequent review still requires sifting through a large number of irrelevant segments.

[0004] Another approach relies on a simple count of failed events to determine anomalies. For example, a fixed time window and a fixed threshold for the number of failures are set; when the number of failures within the window reaches the threshold, an anomaly is identified and recording is initiated. While this method is more targeted than motion detection, it has significant drawbacks: firstly, the fixed threshold cannot adapt to changes in pedestrian density at different times; secondly, factors such as sudden changes in lighting and behavioral differences between registered individuals and strangers are not considered, resulting in an overly simplistic decision-making logic and persistent problems of false positives and false negatives.

[0005] Therefore, there is an urgent need to design an attendance recording control system based on video linkage that triggers abnormal behavior. Summary of the Invention

[0006] The purpose of this invention is to provide a video-linked, abnormal behavior-triggered attendance recording control system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention aims to provide a video-linked, abnormal behavior-triggered attendance recording control system, comprising: The event acquisition and integration unit is used to acquire in real time the recognition success events or recognition failure events generated by personnel on the face recognition attendance device, as well as the timestamps of the two types of events, and to perform weighted accumulation of recognition failure events within a sliding time window to generate a cumulative intensity value. An environmental sensing unit is used to collect environmental parameters of the environment in which the attendance device is located, and the environmental parameters include at least light intensity. An adaptive threshold determination unit is used to dynamically adjust the abnormal behavior trigger threshold according to the environmental parameters, and compare the cumulative intensity value with the abnormal behavior trigger threshold. When the cumulative intensity value exceeds the trigger threshold, it is determined that an abnormal behavior has occurred and the current trigger time is recorded. At the same time, the corresponding abnormal level is output according to the difference between the cumulative intensity value and the threshold. A recording instruction generation unit is used to generate a differentiated control instruction based on the anomaly level at the trigger time. The control instruction includes at least recording quality parameters and storage strategy parameters. The video recording execution unit is used to receive and execute the control command, and simultaneously store the video recordings within the abnormal behavior time period defined by a preset duration extending forward and backward from the trigger time into a designated storage area. An abnormal auxiliary control unit is used to simultaneously perform auxiliary control of evidence locking and ambient lighting enhancement when abnormal behavior is triggered.

[0008] As a further improvement to this technical solution, in the event acquisition and integration unit, the weighted accumulation adopts an exponentially decaying failure accumulation intensity function, as follows: ; In the formula, This represents the cumulative intensity value at the current time t. For the current moment Total number of failure events within the previous time window; ; For the first The weight of each failure event; The attenuation coefficient; For the first The moment of the second failure.

[0009] As a further improvement to this technical solution, the event acquisition and integration unit includes a predictive triggering module. The predictive triggering module is used to acquire the time series of the cumulative intensity value calculated by the exponentially decaying cumulative intensity function of failure; calculate the standard deviation of the prediction error based on the pre-stored historical prediction error; predict the cumulative intensity value after a preset period of time using a first-order exponential smoothing prediction algorithm; and calculate the predicted exceedance probability based on the predicted cumulative intensity value, the current abnormal behavior trigger threshold, and the standard deviation of the prediction error. When the probability exceeds a preset threshold, a pre-buffering instruction is generated and sent to the recording execution unit. After receiving the pre-buffering instruction, the recording execution unit starts the recording buffer in advance and uses a low-resolution, low-frame-rate mode for temporary storage.

[0010] As a further improvement to this technical solution, the environmental perception unit includes a time period identification module. The time period identification module outputs the current time period type based on a real-time clock and a preset attendance time. The time period type includes at least peak period and off-peak period.

[0011] As a further improvement to this technical solution, the adaptive threshold determination unit includes an illumination change compensation module, a time period adaptation module, and an identity association module; The illumination change compensation module is used to detect the change in illumination intensity per unit time. When the change in illumination intensity per unit time exceeds the preset illumination change threshold, the abnormal behavior trigger threshold is switched to the first preset threshold, which is higher than the preset benchmark threshold. The time period adaptive module is used to set the abnormal behavior trigger threshold according to the current time period type output by the time period identification module: during the peak period, the abnormal behavior trigger threshold is switched to a second preset threshold, which is higher than a preset baseline threshold; during the off-peak period, the abnormal behavior trigger threshold is switched to a third preset threshold, which is lower than a preset baseline threshold. The identity association module is used to set a threshold based on whether the identification failure event is associated with the identity of a registered person: when the identification failure event is associated with the identity of a registered person, the abnormal behavior trigger threshold is set to a fourth preset threshold, which is higher than a preset benchmark threshold; when the identification failure event is not associated with an identity, the abnormal behavior trigger threshold is set to a fifth preset threshold, which is lower than a preset benchmark threshold. The preset benchmark threshold is stored in the adaptive threshold determination unit.

[0012] As a further improvement to this technical solution, in the illumination change compensation module, the duration for which the abnormal behavior trigger threshold is switched to the first preset threshold is a preset recovery time. After the preset recovery time has elapsed, the adaptive threshold determination unit resumes using the abnormal behavior trigger threshold before the switch.

[0013] As a further improvement to this technical solution, the anomaly level includes a normal level, a mild anomaly level, and a severe anomaly level, and the recording instruction generation unit specifically comprises: When the anomaly level is mild, a first control command is generated, instructing the recording execution unit to increase the recording frame rate and bit rate to the first preset value respectively, and to add an event tag to the recording of that period. When the anomaly level is severe, a second control command is generated, instructing the recording execution unit to increase the recording frame rate and bit rate to the second preset value respectively, and to store the recording of this period in an independent permanent protection zone; The second preset value is higher than the first preset value.

[0014] As a further improvement to this technical solution, the recording quality parameters include the recording frame rate and the recording bit rate, and the storage strategy parameters include the storage partition identifier; the recording instruction generation unit is specifically configured to generate at least one of the following: a recording frame rate switching instruction, a recording bit rate switching instruction, and a storage partition specification instruction.

[0015] As a further improvement to this technical solution, the anomaly auxiliary control unit includes an evidence locking module and a supplementary lighting control module; The evidence locking module is used to send a locking command to the video recording execution unit when the abnormality level output by the adaptive threshold determination unit is a mild abnormality level or a severe abnormality level. If the event acquisition and integration unit receives a successful identification event, the locking command instructs the video recording execution unit to permanently save video segments of preset durations before and after the successful identification event. The supplementary lighting control module is used to output a pulse width modulation signal to adjust the luminous intensity of the supplementary light when the abnormality level is mild or severe and the ambient light intensity is lower than a preset brightness threshold at the trigger time.

[0016] As a further improvement to this technical solution, the locking instruction sent by the evidence locking module includes a timestamp of the successfully identified event and a locking time range parameter. The video recording execution unit transfers the corresponding video segment from the cyclic overwrite storage area to the permanent protection area according to the locking instruction, and writes event metadata containing at least a timestamp and personnel identification, and this metadata cannot be cleared under normal deletion operations.

[0017] Compared with the prior art, the beneficial effects of the present invention are as follows: This video-linked, abnormal behavior-triggered attendance recording control system achieves sensitive and scene-adaptive detection of abnormal behavior through exponentially decaying failure event accumulation integration and multi-source environment adaptive threshold determination. This effectively avoids false alarms and missed alarms under different time periods, lighting conditions, and personnel identities based on fixed thresholds. At the same time, through predictive pre-buffering and hierarchical differentiated recording linkage, key images can be cached with low overhead before the actual triggering of an anomaly. The recording quality and storage strategy are automatically adjusted according to the anomaly level, thereby significantly reducing storage and computing power consumption while ensuring the integrity of key evidence, and improving the integrity of evidence and the efficiency of system resource utilization. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall process of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example: Please refer to Figure 1 As shown, a video-linked, abnormal behavior-triggered attendance recording control system is provided. This system comprises an event acquisition and integration unit, an environmental perception unit, an adaptive threshold determination unit, a recording instruction generation unit, a recording execution unit, and an abnormal auxiliary control unit working together. Traditional attendance recording systems often employ continuous recording or simple motion detection, which not only wastes storage resources but also struggles to distinguish between normal environmental fluctuations and genuine malicious attempts. This invention transforms failed events into cumulative intensity values ​​that decay exponentially over time and dynamically adjusts the trigger threshold based on environmental parameters. High-quality recording is initiated only when truly needed, while predictive triggering pre-caches potential abnormal footage, thereby significantly reducing storage and computing power consumption while ensuring the integrity of critical evidence.

[0021] The event acquisition and integration unit receives recognition failure events generated by the facial recognition attendance device in real time and attaches a precise timestamp to each event. To avoid the system becoming overly sensitive to occasional failures due to the infinite accumulation of historical events, this unit performs weighted accumulation of failure events within a 60-second sliding time window, generating a cumulative intensity value that represents the concentration of recent failures. Furthermore, the weight of each failure event decays exponentially over time—the later the event occurs, the higher its weight (in the actual implementation, this system simplifies the process, and all event weights are...). With all terms set to 1, the attenuation effect is entirely due to the exponential term. (Provided), the earlier the event, the closer its weight is to zero. This design allows the system to sensitively capture typical abnormal patterns such as multiple consecutive failures in a short period of time, while avoiding prolonged high alert due to a single accidental failure a few hours ago. In specific implementation, this unit uses the following exponentially decaying failure accumulation strength function: ; In the formula, For the current moment The cumulative intensity value; For the current moment Total number of failure events within the previous time window; ; For the first The weight of each failure event is set to 1, and all failure events in this system have a weight of 1. The attenuation coefficient is... ; For the first The time of the next failure event; the default time window is 60 seconds. The system recalculates this value whenever a new failure event occurs. Compared to simple failure counts, the exponential decay model provides a more nuanced depiction of the temporal distribution of failure events: the same number of failures, if concentrated within the last few seconds, will have a significantly higher cumulative intensity value than if scattered over tens of seconds, thus more accurately reflecting the suspiciousness of a "concentrated attack attempt."

[0022] To further capture valuable footage in the early stages of anomalies, a predictive triggering module is integrated within the event acquisition and integration unit. This module uses... The cumulative intensity value is read at intervals of seconds. and will recently A time series is constructed using sampling points (corresponding to the past 5 seconds). The future is predicted using first-order exponential smoothing. The cumulative intensity value after each sampling interval is derived using the following recursive formula: ; In the formula, For the next sampling time (i.e. The cumulative intensity prediction value at (time); This is the smoothed prediction value from the previous sampling time. The smoothing coefficient (0.7 in this embodiment, used to control the weighting of historical and current data); initial time Take the first actual cumulative intensity value Regarding the future The prediction for step (k=4) is performed using an iterative method: (This assumes a constant trend, applicable to short-term forecasts). The final cumulative intensity forecast is obtained after 2 seconds. .

[0023] Predicted values With the current adaptive threshold (That is, the abnormal behavior trigger threshold dynamically adjusted by the adaptive threshold determination unit) is compared, and the standard deviation of the prediction error of the past M sampling points is also referenced. To quantify the uncertainty of the forecast. The forecast error sequence is defined as follows: (in For the past (sampling time) for The actual cumulative intensity value at any given time; for The predicted value at time; then the standard deviation of the prediction error. for: ; ; in For the past The arithmetic mean of the prediction errors at each sampling point is used. Assuming the prediction errors follow a normal distribution with a mean of zero, the probability of exceeding the prediction threshold is calculated based on the distance between the predicted value and the standard deviation of the prediction. for: ; in This represents the cumulative distribution function of the standard normal distribution. It should be noted that this probability calculation implicitly assumes the normality of the prediction error; this system approximates this assumption by continuously updating the error sequence during actual operation. Simultaneously, due to the threshold... As this may change over time, this prediction module uses the current threshold as an approximation, which is engineering-acceptable in short-term predictions.

[0024] Once the probability exceeds the 80% confidence threshold, the system determines that an abnormal behavior is about to occur and immediately sends a pre-buffering instruction to the recording execution unit. Upon receiving this instruction, the recording execution unit initiates a temporary buffer at low resolution (320×240) and a low frame rate (10fps). The buffered data is temporarily stored in a 30-second circular buffer without being written to long-term storage. If the actual cumulative intensity value subsequently exceeds the threshold and triggers an anomaly, the previously buffered low-quality footage will be merged with the high-quality recording after the anomaly as key footage from the early stages of the anomaly, forming a complete chain of evidence. If the prediction is incorrect and no anomaly occurs within 5 seconds, the buffer contents are automatically discarded, with almost no additional storage overhead throughout the process. In this way, the system can intervene in recording hundreds of milliseconds before the anomaly actually occurs, solving the problem of traditional threshold-triggered methods easily losing footage from the early stages of anomalies.

[0025] The environmental sensing unit collects environmental parameters of the environment where the attendance equipment is located, including at least light intensity. It also includes a built-in time period recognition module, which outputs the current time period type based on a real-time clock and a preset attendance schedule (e.g., morning peak 8:00-9:00, evening peak 17:00-18:00, and the rest being off-peak or low-peak periods). The reason for introducing time period information is that during peak hours, the crowds are dense and queues are congested, so recognition failures are often due to obstruction or haste, indicating a lower degree of malicious intent; while during off-peak hours, the crowds are sparse, and a single failure could indicate a probing attack. Using the time period type as the basis for subsequent threshold adjustments can significantly improve the scenario adaptability of the judgment.

[0026] The adaptive threshold determination unit is the core of the system's decision-making. Based on the light intensity and time period type provided by the environmental sensing unit, it dynamically adjusts an abnormal behavior trigger threshold, then compares the cumulative intensity value output by the event acquisition and integration unit with this threshold. When the cumulative intensity value exceeds the threshold, an abnormal behavior is determined to have occurred, and the current trigger time is recorded; simultaneously, the difference between the threshold and the actual intensity value is considered. Output the corresponding anomaly level—the larger the difference, the more severe or concentrated the abnormal behavior, and the higher the anomaly level (divided into three levels: normal, mild anomaly, and severe anomaly), specifically: like The result is classified as normal (no recording control is triggered, or the normal recording status is maintained). like It was determined to be at a mild abnormality level; like It was determined to be of a severe abnormality level; The threshold value (1.0) mentioned above is a configurable parameter, which the system administrator can adjust according to on-site safety requirements.

[0027] To ensure the precision and interpretability of threshold adjustment, a baseline threshold is pre-stored within the unit, and three independently functioning modules are set up: an illumination change compensation module, a time period adaptation module, and an identity association module.

[0028] In this embodiment, the baseline threshold represents the cumulative intensity threshold for the system to determine abnormal behavior under standard environmental conditions (light intensity of 300–500 lux, off-peak hours) and without distinguishing between personnel identities. In this embodiment, the baseline threshold is preferably set to 2.5 (the cumulative intensity value is dimensionless).

[0029] The system has preset several different thresholds for different environmental conditions and event attributes, as follows: The illumination change compensation module uses a first preset threshold. When the change in illumination intensity per unit time exceeds 50 lux / s, it is determined to be a sudden illumination change. At this time, the abnormal behavior trigger threshold switches to the first preset threshold. The first preset threshold is higher than the baseline threshold, which is used to reduce the system's sensitivity to recognition failure events caused by sudden illumination changes. In this embodiment, the first preset threshold is set to 4.0.

[0030] It is important to note that this high threshold state is not permanent; otherwise, the system would become insensitive to genuine abnormal behavior over a prolonged period. Therefore, a preset recovery time is set in the module (5 seconds in this embodiment; this duration can be adjusted via the configuration interface based on the stability of ambient lighting). The timer starts at the switching moment, and after 5 seconds, it automatically reverts to the normal threshold under the current time period and identity conditions.

[0031] Because sudden changes in illumination are usually transient physical phenomena (such as light switching, sunlight being blocked by passing clouds or pedestrians), their interference effect disappears within seconds. An excessively long, lenient threshold window can actually reduce the system's ability to detect simultaneous real attacks. By introducing a configurable recovery time, the system can avoid false triggers caused by short-term illumination fluctuations while quickly recovering to normal sensitivity levels, thus achieving a balance between environmental adaptability and security vigilance. In actual operation, if a new illumination change occurs again within the recovery time, the recovery time will be reset, ensuring that the threshold remains lenient throughout continuous fluctuations and avoiding frequent switching.

[0032] The time-adaptive module uses a second and a third preset threshold. Based on the attendance system, 7:30-9:00 and 17:00-18:30 are preset as peak hours.

[0033] During peak hours, the abnormal behavior trigger threshold switches to a second preset threshold, which is higher than the baseline threshold, to reduce false triggers caused by overcrowding. In this embodiment, the second preset threshold is set to 3.5. During off-peak hours (other than the aforementioned periods), the abnormal behavior trigger threshold reverts to the baseline threshold (2.5) or switches to a third preset threshold. The third preset threshold is lower than the baseline threshold and is used to increase sensitivity to abnormal behavior when there are few people. In this embodiment, the third preset threshold is set to 1.8. The identity association module uses a fourth and a fifth preset threshold. In this system, recognition failure events may be associated with the identity of a registered person. Specifically, during each recognition process, the facial recognition attendance device calculates the similarity between the captured facial image and the registered person templates in the database. When the similarity is higher than a certain threshold (e.g., 60%) but lower than the successful recognition threshold (e.g., 85%), although it is considered a recognition failure, it can be determined that the face corresponds to a registered person (i.e., "high similarity failure"). At this time, the failure event is associated with that person's identity. When a recognition failure event can be associated with a registered person's identity, the system uses the fourth preset threshold. The fourth preset threshold is higher than the baseline threshold, indicating a certain tolerance for consecutive failures by that person (possibly due to personal error). In this embodiment, the fourth preset threshold is set to 3.0.

[0034] When an identification failure event cannot be associated with any registered person (i.e., the similarity is below the association threshold, and the attendance device outputs "unknown person"), the system uses a fifth preset threshold. The fifth preset threshold is lower than the baseline threshold, indicating high sensitivity to repeated failures by strangers, thus preventing malicious probing. In this embodiment, the fifth preset threshold is set to 1.5.

[0035] Priority rules when multiple conditions are met simultaneously: When multiple modules among the illumination change compensation module, time period adaptation module, and identity association module simultaneously meet the conditions, the final abnormal behavior trigger threshold is determined according to the following priority order: Illumination change compensation module > Identity association module > Time period adaptation module. That is, the preset threshold set by the module with the highest priority is used first, regardless of the numerical relationship between the preset thresholds. In this embodiment, this priority order is based on a security strategy: environmental interference caused by sudden changes in illumination requires the most lenient threshold (least sensitive), followed by identity association, and time period adaptation only takes effect when there are no other interferences. For example, if a sudden change in illumination occurs during a peak period, the first preset threshold (4.0) of the illumination change compensation module is used, while the second preset threshold (3.5) of the time period adaptation module is ignored.

[0036] The baseline threshold and all preset thresholds can be remotely configured through the system management software to adapt to the attendance management requirements of different scenarios.

[0037] Once the adaptive threshold determination unit confirms the occurrence of abnormal behavior, the recording instruction generation unit immediately generates differentiated control instructions based on the output abnormality level. These control instructions include at least recording quality parameters and storage strategy parameters; the recording quality parameters include the recording frame rate and recording bitrate, and the storage strategy parameters include storage partition identifiers. Specifically, the recording instruction generation unit is configured to generate at least one of a recording frame rate switching instruction, a recording bitrate switching instruction, and a storage partition specification instruction to adjust the recording quality parameters and storage strategy parameters. For minor anomalies, the instructions require increasing the frame rate from the default 15fps to 30fps and the bitrate from 1Mbps to 3Mbps, and adding searchable event tags (such as timestamps and anomaly levels) to the recordings for that period. For severe anomalies, the instructions further require increasing the frame rate to 60fps and bitrate to 8Mbps, and directly storing the recordings for that period in a separate permanent protection zone—this permanent protection zone uses a non-cyclic overwrite strategy, and only administrators can delete its content. Through this tiered processing, the system ensures that recordings of minor anomalies are sufficient for subsequent verification without consuming excessive storage, while also ensuring that recordings of severe anomalies leave the highest quality legal evidence and prevent cyclic overwriting.

[0038] After receiving control commands, the recording execution unit adjusts the camera's frame rate, bit rate, and the target data to be written to the storage partition. Simultaneously, it needs to define the start and end range of the "abnormal behavior period": In this embodiment, the recording execution unit uses the trigger time as a baseline, extending forward and backward by a preset duration (e.g., 5 seconds forward and 10 seconds backward), identifying and saving the recording segments within this time window as abnormal behavior periods. Extending forward allows playback of the triggers before the anomaly occurred (such as several consecutive failed attempts), while extending backward records subsequent behaviors after the anomaly (such as personnel leaving or security personnel arriving), thus forming a complete chain of evidence.

[0039] The anomaly auxiliary control unit simultaneously executes two auxiliary controls when an anomaly is triggered. The first is evidence locking: when the anomaly level output by the adaptive threshold judgment unit is mild or severe, if the subsequent event acquisition and integration unit receives a successful recognition event (i.e., someone finally succeeds in passing attendance after repeated attempts), then this successful event is very likely to correspond to the executor of the previous abnormal behavior—a legitimate user who finally succeeded after several failures, and the video recording of the moment of success contains clear facial identity information.

[0040] At this point, the evidence locking module immediately sends a locking command to the video recording execution unit. This command carries the timestamp of the successfully identified event and the locking time range parameters (e.g., 5 seconds before and after the successful event). Upon receiving the locking command, the video recording execution unit moves the corresponding time segment from the default circular overwrite storage area to the permanent protection area and writes event metadata containing at least the timestamp and personnel identification. This metadata cannot be erased under normal deletion operations (achieved through file system attributes or anti-deletion database records). This realizes the automatic association from "abnormal behavior" to "identity confirmation," greatly facilitating post-event traceability.

[0041] Secondly, there's the enhanced ambient lighting: At the trigger moment, if the current anomaly level reaches mild or severe, and the ambient light intensity is below a preset brightness threshold (e.g., 30 lux; below this value, the recorded image will be unreadable), the lighting control module outputs a pulse-width modulation (PWM) signal to adjust the intensity of the fill light, quickly raising the image brightness to a suitable level. PWM allows for stepless dimming and avoids flicker interference with the face recognition sensor, ensuring clear and usable evidence footage.

[0042] The entire system operates in a closed loop: under normal conditions, the system maintains the exponentially decaying cumulative intensity value with minimal computational overhead, while the recording execution unit remains in standby or low-frame-rate power-saving mode. Once the cumulative intensity value shows an upward trend, the predictive trigger module preemptively activates the low-quality buffer; if the actual value exceeds a dynamic threshold, it immediately upgrades to high-quality recording with supplemental lighting; if successful identification is subsequently received, it automatically locks key evidence containing the identity. If no threshold is reached, the buffered data is automatically discarded, and the system returns to idle status. This event-driven design ensures complete evidence collection of abnormal behavior while significantly reducing storage usage and power consumption compared to traditional continuous recording solutions.

[0043] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.

Claims

1. A video-linked, abnormal behavior-triggered attendance recording control system, characterized in that, include: The event acquisition and integration unit is used to acquire in real time the recognition success events or recognition failure events generated by personnel on the face recognition attendance device, as well as the timestamps of the two types of events, and to perform weighted accumulation of recognition failure events within a sliding time window to generate a cumulative intensity value. An environmental sensing unit is used to collect environmental parameters of the environment in which the attendance device is located, and the environmental parameters include at least light intensity. An adaptive threshold determination unit is used to dynamically adjust the abnormal behavior trigger threshold according to the environmental parameters, and compare the cumulative intensity value with the abnormal behavior trigger threshold. When the cumulative intensity value exceeds the trigger threshold, it is determined that an abnormal behavior has occurred and the current trigger time is recorded. At the same time, the corresponding abnormal level is output according to the difference between the cumulative intensity value and the threshold. A recording instruction generation unit is used to generate a differentiated control instruction based on the anomaly level at the trigger time. The control instruction includes at least recording quality parameters and storage strategy parameters. The video recording execution unit is used to receive and execute the control command, and simultaneously store the video recordings within the abnormal behavior time period defined by a preset duration extending forward and backward from the trigger time into a designated storage area. An abnormal auxiliary control unit is used to simultaneously perform auxiliary control of evidence locking and ambient lighting enhancement when abnormal behavior is triggered. In the event acquisition and integration unit, the weighted accumulation adopts an exponentially decaying failure accumulation intensity function, as follows: In the formula, This represents the cumulative intensity value at the current time t. For the current moment Total number of failure events within the previous time window; ; For the first The weight of each failure event; The attenuation coefficient; For the first The moment of the second failure event; The environmental perception unit includes a time period recognition module, which outputs the current time period type based on a real-time clock and a preset attendance time. The time period type includes at least peak period and off-peak period. The adaptive threshold determination unit includes an illumination change compensation module, a time period adaptive module, and an identity association module; The illumination change compensation module is used to detect the change in illumination intensity per unit time. When the change in illumination intensity per unit time exceeds the preset illumination change threshold, the abnormal behavior trigger threshold is switched to the first preset threshold, which is higher than the preset benchmark threshold. The time period adaptive module is used to set the abnormal behavior trigger threshold according to the current time period type output by the time period identification module: during the peak period, the abnormal behavior trigger threshold is switched to a second preset threshold, which is higher than a preset baseline threshold; during the off-peak period, the abnormal behavior trigger threshold is switched to a third preset threshold, which is lower than a preset baseline threshold. The identity association module is used to set a threshold based on whether the identification failure event is associated with the identity of a registered person: when the identification failure event is associated with the identity of a registered person, the abnormal behavior trigger threshold is set to a fourth preset threshold, which is higher than a preset benchmark threshold; when the identification failure event is not associated with an identity, the abnormal behavior trigger threshold is set to a fifth preset threshold, which is lower than a preset benchmark threshold. The preset benchmark threshold is stored in the adaptive threshold determination unit; When multiple modules among the illumination change compensation module, time period adaptation module, and identity association module simultaneously meet the conditions, the preset threshold determined by the module with the highest priority is used as the final abnormal behavior trigger threshold, according to the priority order of the illumination change compensation module, identity association module, and time period adaptation module.

2. The video-linked abnormal behavior triggered attendance recording control system according to claim 1, characterized in that: In the illumination change compensation module, the duration for which the abnormal behavior trigger threshold is switched to the first preset threshold is a preset recovery time. After the preset recovery time has elapsed, the adaptive threshold determination unit reverts to using the abnormal behavior trigger threshold before the switch.

3. The video-linked abnormal behavior triggered attendance recording control system according to claim 2, characterized in that: The anomaly levels include normal level, mild anomaly level, and severe anomaly level. The video recording instruction generation unit specifically comprises: When the anomaly level is mild, a first control command is generated, instructing the recording execution unit to increase the recording frame rate and bit rate to the first preset value respectively, and to add an event tag to the recording of that period. When the anomaly level is severe, a second control command is generated, instructing the recording execution unit to increase the recording frame rate and bit rate to the second preset value respectively, and to store the recording of this period in an independent permanent protection zone; The second preset value is higher than the first preset value.

4. The video-linked abnormal behavior triggered attendance recording control system according to claim 3, characterized in that: The event acquisition and integration unit includes a predictive triggering module, which is used to acquire the time series of the cumulative intensity value calculated by the exponentially decaying cumulative intensity function of failure. Calculate the standard deviation of the prediction error based on historical prediction errors; The cumulative intensity value after a preset time period is predicted using a first-order exponential smoothing prediction algorithm. The predicted probability of exceeding the standard is obtained based on the predicted cumulative intensity value, the current abnormal behavior trigger threshold, and the standard deviation of the prediction error. When the probability exceeds a preset threshold, a pre-buffering instruction is generated and sent to the recording execution unit. After receiving the pre-buffering instruction, the recording execution unit starts the recording buffer in advance and uses a low-resolution, low-frame-rate mode for temporary storage.

5. The video-linked abnormal behavior triggered attendance recording control system according to claim 4, characterized in that: The recording quality parameters include the recording frame rate and the recording bit rate, and the storage strategy parameters include the storage partition identifier; the recording instruction generation unit is specifically configured to generate at least one of the following: a recording frame rate switching instruction, a recording bit rate switching instruction, and a storage partition specification instruction.

6. The video-linked abnormal behavior triggered attendance recording control system according to claim 5, characterized in that: The anomaly auxiliary control unit includes an evidence locking module and a supplementary lighting control module; The evidence locking module is used to send a locking command to the video recording execution unit when the abnormality level output by the adaptive threshold determination unit is a mild abnormality level or a severe abnormality level. If the event acquisition and integration unit receives a successful identification event, the locking command instructs the video recording execution unit to permanently save video segments of preset durations before and after the successful identification event. The supplementary lighting control module is used to output a pulse width modulation signal to adjust the luminous intensity of the supplementary light when the abnormality level is mild or severe and the ambient light intensity is lower than a preset brightness threshold at the trigger time.

7. The video-linked abnormal behavior triggered attendance recording control system according to claim 6, characterized in that: The locking instruction sent by the evidence locking module includes a timestamp of the successful identification event and a locking time range parameter.

Images