Video anti-glare interference personnel on-duty identification method, device, equipment and medium

Abstract

The application provides a video anti-glare interference on-duty personnel identification method, device, equipment and medium, the method comprises: acquiring continuous video pictures, extracting human image data and detecting labeled reflection points; combining both in time sequence analysis, generating a reflection suspicious segment; frame by frame comparing the human body contour and the reflection point boundary to determine the misjudgment area, and generating a misjudgment anchor point list; according to the analysis of the ambient light frequency change and the reflection angle distribution, the glare intervention entrance data containing the time window and the spatial range is generated; then the rhythm intervention scheme is generated, the video collection is re-distributed, the collection sequence is limited, the light source is switched, and the focal length is corrected, and the rhythm adjustment and light intervention are adjusted, and the optimized video frame sequence is obtained, and finally the on-duty personnel identification result is output. The method actively adjusts the rhythm and intervenes the light, effectively suppresses the interference of reflection on target detection in strong reflection environment, eliminates the detection box jumping, and outputs stable and accurate identification result.

Description

Technical Field

[0001] This application belongs to the field of video recognition, and in particular relates to a method, device, equipment and medium for identifying personnel on duty using video anti-reflective interference. Background Technology

[0002] With the widespread application of artificial intelligence and computer vision technologies in the security monitoring field, automatic identification of personnel presence based on video analytics is receiving increasing attention. Users need an intelligent monitoring solution that can automatically, in real-time, and accurately determine whether key positions are staffed to improve security management.

[0003] To ensure the stable and reliable operation of a video analytics-based personnel presence recognition system, accurate detection and continuous tracking of human targets in the monitored footage are necessary. One existing technical solution for personnel presence recognition operates as follows: Edge computing nodes and cameras are deployed in the fire control room, receiving continuous video feeds. At the edge computing nodes, target detection algorithms such as YOLOv10 are used to process the video frames, automatically extracting human image data, including recognizing human contours, posture, and coordinates. Based on this human recognition result, the system performs real-time data processing and status determination at the edge, making a judgment on whether a person is on duty without relying on the cloud, and notifying the management terminal when a person is determined to be absent from their post.

[0004] The aforementioned technical solution monitors personnel through computer vision-based human body recognition. However, fire control rooms are typically equipped with numerous metal panels, glass cabinet doors, and monitoring displays—devices with strong reflective properties. These surfaces are prone to forming bright reflective areas when the lighting environment changes, the monitoring screen flickers, or there is interference from external light sources. Because the light intensity distribution, edge gradient, and shape characteristics of the reflective areas are highly similar to the contour features of the human body, computer vision-based human body recognition systems are prone to misidentifying these bright areas as human body edges during detection. Summary of the Invention

[0005] The purpose of this application is to overcome the deficiencies in the prior art and provide a method, device, equipment and medium for video anti-reflective interference personnel on duty identification.

[0006] This application provides a video anti-reflective interference personnel on-duty identification method, including:

[0007] Acquire continuous video feeds from a camera device;

[0008] The continuous video frames are analyzed to extract human image data, and reflection points in the continuous video frames are detected and labeled to obtain reflection point labeling data;

[0009] Based on the human image data and the reflection point annotation data, a comprehensive analysis is performed in chronological order to generate a record of suspected reflective fragments.

[0010] Based on the recorded reflective suspected segments, the human image contour and the boundary of the reflection point are compared frame by frame to determine the starting area of ​​image misjudgment and generate a list of image misjudgment anchor points.

[0011] Based on the list of image misjudgment anchor points, the changes in ambient light frequency and the distribution of reflection angles are analyzed to generate glare intervention entry data that includes the time window and spatial location range that cause misjudgment.

[0012] Based on the glare intervention entry data, a rhythm intervention scheme is generated to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus rebound phase.

[0013] According to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted and lighting is intervened to obtain an optimized video frame sequence;

[0014] Target detection is performed on the optimized video frame sequence, and the on-duty personnel identification result is output.

[0015] Optionally, based on the recorded suspected reflective segments, the human image contour and the boundary of the reflection point are compared frame by frame to determine the starting area of ​​the image misjudgment, including:

[0016] Based on the recorded reflective suspected fragments, a frame-by-frame comparison is performed to generate a basic comparison dataset;

[0017] Based on the aforementioned baseline comparison dataset, the drift path of the human image contour in consecutive frames is calculated;

[0018] Based on the drift path, the relative motion relationship between the reflection point and the edge of the human image is analyzed to determine the starting area of ​​image misjudgment.

[0019] Optionally, a rhythmic intervention scheme is generated for reallocating the video capture order, limiting the timing of light source switching, and correcting the framing rhythm during the camera's focus recovery phase, including:

[0020] Based on the glare intervention entry data, establish a light intervention benchmark table;

[0021] Based on the aforementioned illumination intervention reference table, an exposure rhythm instruction is generated;

[0022] The exposure rhythm command is used to determine the exposure sequence and duration within a stable lighting range to avoid areas of changing lighting, and to reallocate the acquisition order of video frames and limit the timing of light source switching accordingly.

[0023] Optionally, based on the glare intervention entry data, a rhythm intervention scheme is generated for reallocating the video acquisition order, limiting the light source switching time, and correcting the framing rhythm during the camera's focus recovery phase, including:

[0024] Generate instructions for performing viewfinder timing corrections during the camera's focus recovery phase;

[0025] The framing beat correction is used to make the focus recovery process overlap with the stable lighting range.

[0026] Optionally, lighting intervention may be applied to the video acquisition process, including:

[0027] Based on the reflection area determined by the glare intervention entry data, a reverse-phase dark window projection is performed on the reflection area during the video acquisition phase.

[0028] Optionally, according to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted, including:

[0029] Based on the aforementioned rhythm intervention scheme, the target detection cycle and video acquisition sequence are synchronized.

[0030] Optionally, based on the list of misjudged anchor points in the image, the changes in ambient light frequency and the distribution of reflection angles are analyzed, including:

[0031] Based on the list of incorrectly identified anchor points in the image, backtrack the original video frames to extract the image of the reflection point region;

[0032] Based on the extracted image of the reflection point area, calculate the brightness center and direction of the reflected light, and record the reflection angle distribution.

[0033] This application also provides a video anti-reflective interference personnel on-duty identification device, including:

[0034] The acquisition module acquires continuous video footage from the camera device;

[0035] The annotation module analyzes the continuous video frames, extracts human image data, and detects and annotates reflection points in the continuous video frames to obtain reflection point annotation data.

[0036] The analysis module performs a comprehensive analysis based on the human image data and the reflection point annotation data in chronological order to generate a record of suspected reflective fragments.

[0037] The list module compares the human image contour with the boundary of the reflection point frame by frame according to the recorded suspected reflective fragments to determine the starting area of ​​image misjudgment and generate a list of image misjudgment anchor points.

[0038] The intervention module analyzes the changes in ambient light frequency and the distribution of reflection angles based on the list of image misjudgment anchor points, and generates glare intervention entry data that includes the time window and spatial location range that caused the misjudgment.

[0039] The solution module generates a rhythm intervention scheme based on the glare intervention entry data, which is used to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus rebound phase.

[0040] The optimization module adjusts the rhythm and lighting of the video acquisition process according to the rhythm intervention scheme to obtain an optimized video frame sequence.

[0041] The recognition module performs target detection on the optimized video frame sequence and outputs the personnel on-duty recognition result.

[0042] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0043] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the above-described method.

[0044] The beneficial effects of this application are:

[0045] This application provides a method for identifying personnel on duty using video anti-reflective interference, comprising: acquiring continuous video frames from a camera device; analyzing the continuous video frames, extracting human image data, and detecting and labeling reflection points in the continuous video frames to obtain reflection point labeling data; performing comprehensive analysis in chronological order based on the human image data and the reflection point labeling data to generate a record of suspected reflection segments; comparing the human image contour and the reflection point boundary frame by frame based on the record of suspected reflection segments to determine the starting area of ​​image misjudgment and generating a list of image misjudgment anchor points; analyzing the changes in ambient light frequency and the distribution of reflection angles based on the list of image misjudgment anchor points to generate glare intervention entry data containing the time window and spatial location range that cause misjudgment; generating a rhythm intervention scheme based on the glare intervention entry data for reallocating the video acquisition order, limiting the time point of light source switching, and correcting the framing rhythm during the camera's focus recovery phase; adjusting the rhythm of the video acquisition process and intervening in the lighting according to the rhythm intervention scheme to obtain an optimized video frame sequence; performing target detection on the optimized video frame sequence and outputting personnel on duty identification results. This application generates a rhythm intervention scheme to actively adjust the rhythm and lighting of the video acquisition process, thereby effectively suppressing the interference of reflection on the target detection algorithm in complex and highly reflective lighting environments, eliminating the jumping of the detection box between the human body and the reflective area, and thus outputting stable and accurate personnel on-duty recognition results. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the personnel on-duty identification process for video anti-reflective interference in this application;

[0047] Figure 2 This is a schematic diagram of the video anti-reflective interference personnel on-duty identification device in this application. Detailed Implementation

[0048] Exemplary embodiments of the present disclosure will now be provided in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it is to be understood that various forms of implementation of the present disclosure are intended and should not be limited to the embodiments set forth herein. Rather, the embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0049] This application provides a video anti-reflective interference personnel on-duty identification method, applied in the fields of fire control room duty monitoring and artificial intelligence monitoring technology. Specifically, it relates to an edge computing-based intelligent monitoring method for personnel leaving their posts in fire control rooms. This method addresses the problem that in fire control rooms, the surfaces of equipment and devices with strong reflective properties, such as metal panels, glass cabinet doors, and monitoring displays, easily form bright reflective areas when the lighting environment changes, the monitoring screen flickers, or external light source interference occurs. Because the light intensity distribution, edge gradient, and shape characteristics of the reflective areas are highly similar to human contour features, computer vision-based human recognition systems are prone to misidentifying these bright areas as human body edges during detection.

[0050] The aforementioned misidentification causes the detection frame to frequently jump between real human bodies and false reflection areas, resulting in unstable identification results over time. This is especially true in dynamic scenarios such as personnel movement, bending, or turning, where instantaneous changes in light reflection further amplify the probability of misjudgment, causing the system to identify reflected signals as independent targets, leading to multiple detection frames or false on-duty results. Once this technical problem occurs, it results in distorted off-duty detection, causing the monitoring system to fail to accurately identify unattended states at critical moments, thus affecting the reliability of safety management in the fire control room.

[0051] This application, based on edge computing for intelligent monitoring of personnel absence in fire control rooms, refers to the deployment of edge devices and cameras with computing capabilities within the fire control room. Human body recognition algorithms are used to analyze video footage in real time to determine whether on-duty personnel are within their designated workstations. Utilizing a deep learning-based human detection and recognition model, the system automatically identifies human contours, postures, and position coordinates, completing data processing and status determination on the edge in real time. When personnel are detected as having been absent from the designated area for an extended period or having left their posts, the edge devices can trigger alarm pushes without relying on the cloud, notifying the management to take appropriate action.

[0052] This application enables continuous sensing and automatic identification of the on-duty status of personnel, improves the real-time safety management level of the fire control room, avoids omissions in manual inspections, and ensures that key positions are manned and the monitoring system continues to operate.

[0053] like Figure 1 As shown, the video anti-reflective personnel on-duty identification described in this application includes:

[0054] S101. Acquire continuous video footage from the camera device.

[0055] It receives the continuous video signal output from the camera device installed in the fire control room to obtain continuous monitoring footage.

[0056] The video signal continuously output from the camera device is input to the edge computing node in a continuous frame manner. The edge computing node receives the video frames sequentially according to the set time interval and performs timing adjustment to ensure that the sampling time interval of each frame is consistent with the previous frame. After the timing adjustment is completed, the edge computing node calculates the overall brightness value of the image frame by frame to determine the current lighting status and screen light source change status in the fire control room.

[0057] To address local brightness fluctuations in the monitoring image caused by light flickering, display switching, or external light source interference, the edge computing node balances the inter-frame brightness change rate, ensuring that the light intensity distribution between adjacent frames remains continuously changing. This provides a stable reference frame sequence, guaranteeing the continuity and stability of the image at the illumination level.

[0058] For example, in an environment with periodically flashing emergency indicator lights, this processing can smooth out sudden changes in brightness and prevent information loss due to overexposure or underexposure of a single frame.

[0059] S102. Analyze the continuous video frame, extract human image data, and detect and label the reflection points in the continuous video frame to obtain reflection point labeling data.

[0060] Human targets are detected in flicker-free video footage, and abnormally bright reflective areas are identified simultaneously. Specifically, personnel areas within the fire control room are identified from each frame. Edge computing nodes divide the viewing area of ​​the fixed-position cameras within the fire control room into several monitoring blocks, including the control console, operator chair, area in front of the monitoring screen, passageways, and workbenches.

[0061] By comprehensively analyzing the pixel grayscale distribution, color features, and edge contours within these monitoring blocks, image regions with human morphological characteristics are identified. The human image extraction process includes three steps:

[0062] First, perform contour boundary detection on the pixel distribution in each frame of the image to determine contour line segments that conform to human features.

[0063] Second, the detected continuous contour segments are combined into an overall contour area to ensure that the shape of the area is consistent with the proportions of the human body.

[0064] Third, the image data corresponding to the identified contour areas are extracted and their coordinate positions and orientations are recorded in the form of frame numbers and timestamps.

[0065] To maintain the temporal continuity of human image data, edge computing nodes match the coordinates of the same human image in adjacent frames, thus establishing a correlation between the human position and motion trajectory between consecutive frames, thereby obtaining a complete human image dataset that is continuous in both time and space.

[0066] After extracting human image data, the edge computing nodes detect and label abnormal brightness points in the same frame of image.

[0067] To achieve accurate identification, edge computing nodes compare the light intensity distribution map of the current frame with the average light intensity distribution in the reference frame sequence point by point, identifying areas where the brightness variation exceeds the normal range of human body light reflection. These areas typically appear on metal panels, glass cabinet doors, or display screens. Edge computing nodes determine whether these areas intersect or are close to the human body image contour based on their brightness gradient, shape features, and spatial location. When an abnormal brightness area is detected to overlap with the human body contour boundary or be adjacent to it within a short distance, that area is identified as a reflection point.

[0068] For each reflection point, the edge computing node extracts its spatial coordinates, brightness intensity value, reflection area size, and duration (number of frames), and records this information as reflection point annotation data. Simultaneously, the edge computing node assigns a unique identifier to each reflection point and associates it with the human image data in the corresponding frame, thus binding the reflection point and the human image in both time and space dimensions.

[0069] The location, duration, and light intensity changes of all reflection points in the video sequence are fully recorded. For example, when a staff member approaches a bright metal control panel, their reflection may be identified by the system as an independent area of ​​abnormal brightness and marked as a reflection point.

[0070] S103. Based on the human image data and the reflection point annotation data, a comprehensive analysis is performed in chronological order to generate a record of suspected reflective fragments.

[0071] The system correlates and analyzes human images and reflection point information over time to identify continuous video segments that may cause recognition interference due to reflections. Specifically, nodes perform comprehensive analysis of human images and reflection point information in consecutive frames in chronological order to generate records of suspected reflection segments. Edge computing nodes use frame numbers as indexes to extract the scene segments in consecutive frames that simultaneously contain human images and reflection points, forming a set of temporally continuous segments.

[0072] For each segment, the edge computing node records its start frame number, end frame number, corresponding human body position coordinates, reflection point spatial coordinates, light intensity change trend, reflection duration, and overlap ratio between the human body and the reflection point. All extracted segments are arranged in chronological order to form a complete record of suspected reflective segments.

[0073] The recording of suspected reflective fragments preserves the correspondence between human image data and reflective point feature data, reflecting the temporal and spatial distribution characteristics of reflective interference.

[0074] Through this record, edge computing nodes can track the specific location and time range of reflection interference in the fire control room, providing basic data support for subsequent identification of the initiation area of ​​reflection misjudgment and interference suppression. A high risk of misjudgment only exists when human images and reflection points coexist continuously in time and space or exhibit specific movement relationships; this record is precisely for locating these high-risk periods.

[0075] Furthermore, during the process of generating suspected reflective segments, the video signals continuously output by the camera installed in the fire control room are input to the edge computing node. Video frames are received and time-series adjusted according to a set time interval. A stable reference frame sequence is obtained through inter-frame brightness change rate balancing processing.

[0076] Based on the suspected reflective fragment records, the human image contour and the boundary of the reflection point are compared frame by frame to determine the starting area of ​​image misjudgment, including: performing frame-by-frame comparison based on the suspected reflective fragment records to generate a basic comparison dataset; calculating the drift path of the human image contour in consecutive frames based on the basic comparison dataset; and analyzing the relative motion relationship between the reflection point and the edge of the human image based on the drift path to determine the starting area of ​​image misjudgment.

[0077] In practice, after obtaining records of suspected reflective segments containing human image data and reflective point annotation data, each frame is read sequentially. Each frame contains a corresponding frame number, timestamp, a set of human image contour coordinates, and a set of reflective point spatial coordinates. The timestamps are used to correct the order of consecutive frames, ensuring that the video frames remain continuous in the time dimension.

[0078] S104. Based on the recorded suspected reflective segments, compare the human image contour with the boundary of the reflection point frame by frame to determine the starting area of ​​image misjudgment and generate a list of image misjudgment anchor points.

[0079] The image content of each frame is then decomposed, the human body image contour coordinate set is extracted separately, and the spatial coordinate set of reflection points in the same frame is independently labeled.

[0080] During the comparison process, for each frame of image, all reflection point positions are compared one by one with the edge points in the human contour coordinate set according to pixel spatial coordinates, recording the spatial distance, relative direction, and positional relationship between the two. When the reflection point at the same position remains stable in brightness across multiple consecutive frames, while the human contour edge shifts near the same position, the change in distance between the reflection point and the adjacent contour edge is recorded as comparison sequence data. Through this process, a basic comparison dataset covering the entire suspected reflective segment is formed, which fully reflects the relative movement of the human contour and reflection points in time and space. After establishing the basic comparison dataset, the positional changes of the human image contour edges in consecutive frames are processed to obtain the drift path of the human image contour. The specific process is as follows:

[0081] For each time point in the comparison dataset, a difference analysis is performed between the set of coordinate points of the human body contour edge and the set of coordinate points at the same position in the previous frame, calculating the offset direction and offset distance of the contour points. The offset directions of all frames are sorted chronologically, and the spatial movement trajectories of the contour points are concatenated into a continuous path according to time sequence. This path records the dynamic positional changes of the human body image in the temporal dimension and shows the trend of human body edge movement in the spatial dimension.

[0082] Simultaneously, to ensure the continuity of the drift path, the position data of all reflection points within the same time window are recorded synchronously with the human contour drift path, so that the relative position of the corresponding reflection point can be obtained at each time node of the path. When the contour frame continuously moves towards the reflection point and contacts or overlaps, this time period is recorded as the key comparison interval, providing direct evidence for determining the misjudgment starting area. Based on the spatial recording results of the image contour drift path, the relative motion relationship between the reflection point and the edge of the human image in consecutive frames is analyzed to determine the location and range of the image misjudgment starting area. The specific steps are as follows:

[0083] In the comparison data, the time frame in which the human figure's outline first spatially overlaps with or approaches a predetermined distance threshold with the reflection point is identified. The coordinate region of the human figure's image edge in this frame is extracted and marked as the center position of the misjudgment starting region. Based on this, the brightness distribution of this time frame and several adjacent frames are compared and analyzed to calculate the degree of matching between the direction of brightness gradient change in this region and the direction of brightness change of the reflection point. When the directions of change of both are consistent and the rates of change are similar, this region is confirmed as the misjudgment starting region affected by reflection.

[0084] Subsequently, based on the spatial displacement of consecutive frames, the timeframe is traced back to the point when the human image contour begins to deviate from its normal trajectory, and then extended backward to the point when the human image contour returns to a stable position. The duration of the misjudged region is determined using all frames within this time interval as boundaries. In the spatial dimension, the drift distance of the human image edge in the horizontal and vertical directions is determined based on the center point of the misjudged starting region, thus forming a misjudged region dataset that includes both time frame intervals and spatial coordinate intervals.

[0085] Based on the analysis results of the initial region of misjudgment, a list is generated that accurately locates the key information (time, space, direction) of each misjudgment event.

[0086] After locating the misjudged areas, the center point data of all the starting areas of the misjudgments are integrated into a list of image misjudgment anchor points in chronological order. Based on the misjudged area dataset, the center point data of each starting area of ​​the misjudgments are integrated to generate a list of image misjudgment anchor points that includes timestamps, spatial coordinates, human contour drift direction, and reflection point positions.

[0087] Each anchor point contains the frame number of the misjudged starting area, timestamp, spatial coordinates, human body contour drift direction, reflection point position, light intensity change trend, and misjudgment duration.

[0088] To facilitate subsequent calls, all anchor points are numbered sequentially along the timeline, and adjacent anchor points are associated according to their order of occurrence based on the temporal continuity of the misjudged areas. When the same reflection point triggers multiple misjudgments at different time periods, its corresponding anchor points are grouped together, and the corresponding relationships are recorded in the list.

[0089] Once the list of image misjudgment anchor points is formed, it fully reflects the temporal patterns and spatial distribution of human images misjudged due to reflective interference in the fire control room monitoring scenario, providing an accurate spatial coordinate reference and temporal positioning basis for subsequent ambient light intervention, exposure rhythm adjustment and acquisition timing control.

[0090] For example, the list might record information such as "At timestamp T1, the reflection point located at coordinates (X1, Y1) caused the human silhouette to drift D pixels northeast."

[0091] S105. Based on the list of image misjudgment anchor points, analyze the changes in ambient light frequency and the distribution of reflection angles to generate glare intervention entry data that includes the time window and spatial location range that cause misjudgment.

[0092] Specifically, by tracing back to the original video environment at the time of the misjudgment, analyzing the characteristics of light fluctuations and reflected light paths, the root cause of the misjudgment and its scope of influence can be identified.

[0093] Specifically, based on the timestamps, frame numbers, spatial coordinates, brightness change trends of reflection points, and human image contour drift directions recorded in the list of misjudged anchor points, the original video data is traced back one by one. According to the time information corresponding to each anchor point, the original frame sequence of the continuous time period before and after the anchor point is extracted from the video storage sequence to ensure that the time covers the entire process before and after the misjudged event.

[0094] The extracted video frames are rearranged chronologically to form a continuous video segment. Edge computing nodes perform brightness analysis on each frame of this segment, extracting the average brightness value and local brightness distribution information for the entire frame, and calculating the illumination intensity variation curves separately for the human image area and reflection point area. The illumination intensity variation curves are plotted with the frame number on the horizontal axis and the brightness value on the vertical axis, thus reflecting the dynamic changes in illumination during video capture.

[0095] For each anchor point, a time-continuous illumination change database is established for the corresponding video segment, providing fundamental data for analyzing ambient light frequency fluctuations and light source status. After obtaining the time-continuous illumination change database, the ambient light frequency changes during the video acquisition phase are further analyzed. The lighting system in a fire control room typically includes main lighting, warning lights, emergency indicator lights, and multiple display terminal screens. These light sources exhibit periodic fluctuations and instantaneous brightness changes during operation. Edge computing nodes extract periodic features from the average brightness curve of consecutive frames to identify the alternating periods of light intensity over time. When a regular brightness change is detected in a light source over time, the illumination frequency, fluctuation period, and duration of that light source are recorded. For illumination changes within the time period of the misjudged anchor point, the occurrence times of light source switching, overlapping light sources, or flickering phenomena are analyzed, and the frame numbers and brightness change amplitudes of these phenomena are recorded. By comparing the time difference between the anchor point occurrence time and the illumination change time, the temporal correlation between misjudgment and light frequency changes is determined. When a misjudged event occurs simultaneously with a fluctuation in illumination, the time interval is designated as an optical frequency anomaly interval, and its start and end frame numbers, fluctuation period, and brightness difference are recorded to provide a time reference for subsequent spatial reflection distribution analysis.

[0096] Based on the image misjudgment anchor point list, the ambient light frequency variation and reflection angle distribution are analyzed, including: based on the image misjudgment anchor point list, the original video frames are traced back to extract the reflection point area image; based on the extracted reflection point area image, the brightness center and direction of the reflected light are calculated, and the reflection angle distribution is recorded.

[0097] In practice, after obtaining the optical frequency anomaly interval, a retrospective analysis is performed on the brightness flicker and reflection angle distribution of the monitored image. Edge computing nodes extract the image of the area containing the reflection point from the original video frames according to the frame number within the optical frequency anomaly interval, and calculate the brightness center position of this area in consecutive frames. Based on the trajectory of the brightness center's position change in space, the trend of reflected light direction change is determined. If the brightness center of the reflected light moves along a fixed direction in consecutive frames, it indicates that the incident angle between the reflection source and the camera has changed. The trajectory of the brightness center of the reflected light is plotted as a path line in the image coordinate system, and the path's start point, end point, and direction vector are recorded.

[0098] Subsequently, the human image contour drift path and the reflected light path are superimposed and compared to analyze their intersection relationship. When the two paths overlap or approach each other in spatial coordinates and have the same time frame number, the area is identified as a false trigger point. To determine the distribution range of the reflection angle, the diffusion area of ​​the reflected light and the degree of edge brightness attenuation in each frame are calculated to determine the incident direction and reflection range of the reflected light.

[0099] This range is recorded in rectangular coordinates to form a spatial data set containing spatial boundary coordinates, time frame sequences, and directions of reflection angle changes. After obtaining the optical frequency anomaly range and spatial reflection angle data, the correspondence between the two is analyzed to generate glare intervention entry data.

[0100] Edge computing nodes aggregate the temporal information, spatial location, light source type, illumination fluctuation characteristics, reflection angle direction, and human image contour drift trajectory of each misjudged event to form a complete data entry. Each data entry includes the event number, start time frame number, end time frame number, spatial coordinate region where the misjudgment occurred, reflected light direction vector, light source frequency parameters, brightness fluctuation curve, and distance change trend between the reflection point and the human image.

[0101] All data entries are arranged chronologically to form the glare intervention entry dataset. This dataset stores the time window, spatial location range, illumination characteristics, and reflection angle information for each misjudged event, and provides complete input for subsequent exposure rhythm adjustments, framing corrections, and illumination synchronization control.

[0102] After the data is generated, the edge computing nodes use it as the core reference for subsequent intervention steps, enabling the system to perform targeted and rhythmic interventions for specific periods of light change and spatial areas, thereby maintaining the stability and continuity of human image recognition in complex lighting environments.

[0103] By combining optical frequency anomaly intervals and reflection angle data, and summarizing time information, spatial coordinates, light source type, and brightness fluctuation curves, glare intervention entry data containing time windows and spatial location ranges is generated.

[0104] Furthermore, during the generation of the glare intervention entry dataset, the direction of reflection angle change within the optical frequency anomaly interval is compared with the human image contour drift trajectory in time. When the time frame numbers of the two are consistent and their spatial positions overlap, the area is identified as the misjudgment trigger point, and the corresponding time interval and spatial coordinates are used as key positioning parameters in the glare intervention entry dataset.

[0105] S106. Based on the glare intervention entry data, generate a rhythm intervention scheme for reallocating the video acquisition order, limiting the light source switching time point, and correcting the framing rhythm during the camera's focus rebound phase.

[0106] Based on the analysis of the causes of light interference, a detailed plan was developed to control the working rhythm of video acquisition equipment (cameras) in order to avoid or counteract the identified light interference.

[0107] Therefore, a light intervention benchmark table was constructed based on the time window, spatial coordinates, light source frequency, reflection angle direction, light fluctuation curve and human image trajectory information contained in the glare intervention entry data.

[0108] The specific process is as follows: For each glare intervention data entry, extract its corresponding time period and light intensity change information, and arrange these time periods in order of frame number on a unified time axis. The time axis marks the start time, end time, brightness change cycle, and fluctuation range corresponding to each light source.

[0109] Subsequently, these time periods were mapped to spatial coordinates, and the correspondence between the light source illumination area, the reflection area, and the human image activity area was recorded in a table. For different light source types, the illumination direction, illumination intensity, fluctuation frequency, and reflection angle information were recorded, and the boundary range of the illumination influence area was determined based on the relative position of the light source and the camera.

[0110] The lighting intervention reference table formed in this way fully demonstrates the changing patterns of the light source in the temporal and spatial dimensions, and provides a temporal reference and spatial reference for the design of exposure rhythm.

[0111] A rhythm intervention scheme is generated for reallocating the video capture order, limiting the light source switching time, and correcting the framing rhythm during the camera's focus recovery phase. This includes: establishing a lighting intervention reference table based on the glare intervention entry data; and generating an exposure rhythm instruction based on the lighting intervention reference table. The exposure rhythm instruction is used to determine the exposure sequence and duration within a stable lighting range to avoid lighting variation ranges, and accordingly reallocates the video capture order and limits the light source switching time.

[0112] In practice, after the lighting intervention reference table is established, exposure rhythm instructions are generated based on the time axis and light source characteristic data in the table. These instructions determine the exposure sequence and duration allocation for the camera at different lighting stages. The specific process is as follows:

[0113] Analyze the fluctuation patterns of the light source over time to identify stable and variable illumination ranges. Avoid sampling during periods of variable illumination by adjusting the exposure time to within the stable illumination range.

[0114] The exposure sequence and interval are set according to the time periods during which the light sources alternate. When the exposure times of two light sources overlap, the exposure sequence is delayed so that the camera pauses capturing the image at the midpoint between the two light source transitions, preventing sudden changes in image brightness caused by abrupt changes in illumination.

[0115] For situations where the light source flicker period is short, shorten the duration of a single exposure so that each acquisition covers the low fluctuation part of the illumination change, thereby obtaining an image with stable brightness.

[0116] The exposure rhythm command also adjusts the exposure rhythm according to the direction of the reflection angle. When the direction of light incidence is consistent with the camera's framing direction, the camera captures the image during a period of low reflected light intensity by controlling the initial delay of the exposure, thereby reducing reflection interference.

[0117] Through the above operations, an exposure rhythm instruction set is formed, which includes time segments, exposure start and end times, sampling intervals, and exposure duration, providing a time frame for the reallocation of the acquisition sequence.

[0118] After generating the exposure rhythm command, the video frame acquisition order is reallocated, and the light source switching time is specified. Specifically, the implementation is as follows:

[0119] Based on the spatial coordinate information of the light source distribution and reflection area in the illumination intervention benchmark table, the camera's acquisition task is divided into multiple areas, each corresponding to the time period of different light sources.

[0120] Based on the time allocation given in the exposure rhythm instructions, the acquisition order of these areas is rearranged according to the degree of light interference. Areas with stable lighting are acquired first, while areas with frequent lighting fluctuations are acquired later.

[0121] During the adjustment of the acquisition sequence, ensure that the images acquired within the same time period are not affected by the overlap of multiple light sources.

[0122] To further control exposure interference during light source switching, the timing of light source switching is limited. This limitation is achieved by setting buffer times before and after the light source switching. When a light source is detected to be about to turn off, the start time of the next frame acquisition is delayed, ensuring the camera begins acquisition only after the light source is completely off. When a light source is about to turn on, the current frame acquisition is completed ahead of schedule, and the inter-frame interval is extended to avoid the instantaneous brightness increase when the light source turns on.

[0123] When multiple light sources switch alternately, a fixed switching interval is preset on the timeline to maintain a stable rhythm between light source switching, preventing illumination fluctuations from concentrating within the same time window. After reallocating the acquisition sequence and limiting the light source switching time points, the time distribution of video acquisition and illumination changes are coordinated and matched.

[0124] A rhythmic intervention scheme is generated for reallocating the video acquisition order, limiting the time point of light source switching, and correcting the framing rhythm during the camera's focus recovery phase. This includes: generating instructions for performing framing rhythm correction during the camera's focus recovery phase; wherein the framing rhythm correction is used to make the focus recovery process overlap with the stable lighting interval.

[0125] Specifically, after completing the exposure rhythm instruction construction and acquisition sequence adjustment, the focus recovery phase during the camera's framing process is corrected to create a complete rhythm intervention scheme. The focus recovery phase is the focus restoration process that occurs when the camera is autofocusing, zooming, or adjusting the framing angle; during this phase, the image is easily affected by lighting fluctuations. Therefore, the focus recovery process is incorporated into the rhythm control system. The specific operation is as follows:

[0126] The time distribution of the focus recovery phase is determined based on the exposure rhythm command and matched with the light fluctuation range. If the focus recovery and light change occur simultaneously, the focus adjustment is delayed by extending the acquisition end time of the previous frame, so that the focus recovery phase overlaps with the stable light range, thereby avoiding focusing when the light is unstable.

[0127] Meanwhile, a silent sampling period is inserted after the focus rebound ends, so that the camera does not capture images during the short interval, providing a buffer time for focus stabilization and lighting balance.

[0128] After each exposure, the framing time of the next frame is adjusted to keep the acquisition cycle synchronized with the cycle of light source brightness changes.

[0129] By correcting the framing rhythm in this way, the camera can always frame the shot in a relatively stable brightness stage in scenes with frequent lighting fluctuations, ensuring the brightness consistency and edge sharpness of continuous images.

[0130] Once the rhythm intervention scheme is formed, it includes four aspects: exposure timing planning, acquisition sequence allocation, light source switching time limit, and framing rhythm correction, providing a complete time control framework for subsequent rhythm adjustment and illumination suppression.

[0131] When performing viewfinder beat correction, the matching between the focus recovery phase and the illumination fluctuation range is achieved by extending the end time of the previous frame acquisition and inserting a silent sampling period. This allows the camera to complete focus recovery during the stable illumination phase and maintain a stable illumination state before the next frame starts, thereby ensuring the brightness consistency and edge sharpness of continuous images.

[0132] S107. According to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted and the lighting is intervened to obtain an optimized video frame sequence.

[0133] By implementing the aforementioned rhythm intervention scheme, a set of high-quality video frames with minimal reflection interference is obtained through precise control of the acquisition timing and the application of active illumination suppression. The YOLOv10 target detection cycle is synchronized with the video acquisition timing in the edge computing nodes according to the rhythm intervention scheme. The specific process is as follows:

[0134] A unified time control benchmark is established in the edge computing nodes to correspond the time interval of video acquisition frames with the execution time of the detection cycle, so that acquisition and detection maintain a strict synchronization relationship on the time axis.

[0135] By segmenting the timeline, the detection window corresponding to each acquisition stage is clearly defined, ensuring that each frame can immediately enter the detection and processing stage after acquisition.

[0136] When fluctuations in lighting or changes in light sources cause changes in screen brightness, the time control benchmark automatically delays the start time of the detection cycle, so that the detection process avoids periods of unstable lighting.

[0137] Meanwhile, during the focus recovery phase, the detection trigger time is delayed to ensure image stability before executing the detection task. Through this precise time alignment, video acquisition and target detection form a synchronous execution mechanism, ensuring that each detection is performed under stable lighting and clear image conditions.

[0138] According to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted, including:

[0139] Based on the aforementioned rhythm intervention scheme, the target detection cycle and video acquisition sequence are synchronized.

[0140] After the time synchronization of acquisition and detection is completed, the video acquisition process is rhythmically adjusted according to the rhythm control parameters in the rhythm intervention scheme in order to establish an acquisition pattern that adapts to changes in lighting.

[0141] Specifically, the video acquisition cycle is divided into a stable range and a fluctuating range according to the light source status. A constant acquisition frequency is maintained in the stable range, while the inter-frame interval is dynamically adjusted in the fluctuating range.

[0142] When the light source enters the phase of increasing brightness, the acquisition interval is extended so that the camera avoids exposure at the moment of sudden change in light; when the light source brightness gradually decreases, the acquisition interval is shortened to increase the number of samples, thereby smoothing the brightness change process.

[0143] During periods of alternating illumination from multiple light sources, the acquisition rhythm is allocated in a layered structure, prioritizing the acquisition of areas controlled by the main light source and delaying the acquisition time of areas affected by secondary light sources.

[0144] To prevent abrupt changes between adjacent lighting phases, a buffered acquisition frame is set at the transition moment of light source change, allowing the image to gradually return to balance during the brightness change transition period. Through the above adjustments, the time distribution of the video acquisition process is kept consistent with the trend of lighting changes, thereby avoiding interference from sudden brightness changes on the detection image.

[0145] Lighting intervention during video acquisition includes:

[0146] Based on the reflection area determined by the glare intervention entry data, a reverse-phase dark window projection is performed on the reflection area during the video acquisition phase.

[0147] Specifically, based on rhythmic adjustment, inverse dark window projection is performed to control brightness interference in reflective areas during the video acquisition stage.

[0148] The inverted dark window is achieved by suppressing directional illumination in a specific area during the video capture period. The implementation method is as follows: the spatial coverage of the inverted dark window is determined based on the coordinates of the reflection area, the direction of the reflection angle, and the changes in illumination intensity recorded in the glare intervention entry data. Subsequently, before the camera exposure begins, an illumination reduction signal is applied to the reflection area, ensuring that the brightness of this area remains constant during exposure and is unaffected by changes in external light sources. For reflection points appearing at different time periods, the edge computing nodes adjust the projection area of ​​the inverted dark window in real time according to their position changes, ensuring that the illumination control range always covers the reflection interference area.

[0149] When the camera's focal length rebounds, the size and position of the inverted dark window are readjusted according to the focal length to ensure that the masking effect remains effective.

[0150] The duration of the inverted dark window is synchronized with the exposure rhythm, precisely opening and closing in each exposure cycle, thereby creating a light suppression rhythm throughout the video capture process and reducing bright interference from reflective areas.

[0151] While performing the reverse dark window projection, soft light scattering guidance is applied to improve the overall lighting balance of the image and further reduce the reflection effect. Soft light scattering guidance makes the lighting transition of the image more natural by adjusting the incident angle and distribution density of the light. Specifically, it is achieved by selecting an auxiliary lighting angle corresponding to the camera's viewpoint based on the relationship between the light source direction and the reflection angle distribution in the rhythm intervention scheme, forming a multi-angle scattered light field around the reflection area.

[0152] The scattered light field generates a uniformly distributed illumination layer in space, dispersing high-intensity light to the surrounding area and reducing the light intensity concentration at the central reflection point.

[0153] To ensure the continuous effect of soft light scattering, the intensity and direction of the scattered light are adjusted in real time during the light source switching period to keep it synchronized with the brightness changes of the main light source.

[0154] When the light source is in a stable phase, soft light scattering maintains a constant brightness distribution, reducing illumination differences in the image and thus preventing strong contrast between the edges of human images and reflective areas. The soft light scattering guidance and the control cycle of the inverted dark window are consistent, and the two work together to maintain a stable state in terms of illumination balance in the monitoring image, providing continuous and uniform image input for subsequent detection.

[0155] Under stable operation guided by inverted dark window projection and soft light scattering, a silent sampling period is set to ensure coordination between the detection cycle and video acquisition. The silent sampling period is set between two consecutive video acquisitions to maintain stable image brightness and complete focus recovery after exposure. The specific process is as follows:

[0156] Based on the time allocation in the rhythm intervention plan, a fixed silent interval is inserted after each exposure to prevent the camera from performing acquisition tasks during this period.

[0157] The duration of the silent sampling period is matched with the light source change cycle to ensure that illumination fluctuations have returned to a stable state before the next frame is captured. During the silent period, the edge computing node keeps the image data cache unchanged and performs illumination equalization and edge smoothing processing on the previously captured image to prepare balanced illumination conditions for the next frame capture.

[0158] After the silent period ends, the data collection process is restarted, thus forming a cyclical pattern of data collection, silence, and detection in terms of time rhythm.

[0159] Through this temporal structure, the system can continuously suppress brightness abrupt changes caused by light source flicker or reflection interference, keeping the detection results stable in a continuous time series. After the above rhythmic adjustment and illumination intervention, a set of optimized video frame sequences is finally obtained. These frames have uniform brightness, clear outlines, and are minimally affected by reflection interference.

[0160] S108. Perform target detection on the optimized video frame sequence and output the personnel on duty identification result.

[0161] An object detection algorithm is applied to the optimized video frame sequence to accurately identify whether personnel are on duty.

[0162] After the silent sampling period ends, the target detection task is started synchronously. The YOLOv10 target detection algorithm is used to perform human body recognition on the optimized video frame sequence. Since glare interference in the image has been effectively suppressed through the aforementioned steps, the outline of the human figure in the recognition image is clear, and the detection box position is continuous without fluctuation. The algorithm can stably and accurately identify personnel within the preset workstation area of ​​the fire control room and output their position coordinates and posture information.

[0163] By continuously analyzing the recognition results of multiple frames, the system can reliably determine whether personnel are on duty. When it detects that personnel have not appeared in the designated area or have left their post for an extended period, an alarm can be triggered. The output on-duty recognition results maintain consistency and reliability across multiple video frames, thereby achieving accurate identification and continuous tracking of personnel absence status, significantly improving the safety and reliability of fire control room monitoring.

[0164] This application achieves rhythmic coordination between video acquisition and target detection cycles in edge computing nodes, dynamically synchronizing the image acquisition sequence with changes in illumination. This maintains image input stability even under conditions of frequent light source flickering and brightness fluctuations in reflective areas. Through the synergistic effect of exposure rhythm control, light source switching limitations, and focus rebound correction, video frames maintain balanced brightness and clear outlines even in environments with lighting interference. This effectively eliminates the problem of human detection box jumps caused by reflective areas, improving the continuity and reliability of personnel on-duty recognition results.

[0165] This application employs a combination of reverse-phase dark window projection, soft light scattering guidance, and multi-layered illumination intervention during silent sampling periods to proactively reduce the interference from highly reflective areas during video acquisition, effectively separating human images from reflective areas in terms of spatial brightness distribution. This method reduces the misleading effect of reflected light signals on the detection model's recognition boundaries, ensuring stable detection accuracy of human targets in complex lighting environments. It enables accurate identification and continuous tracking of personnel off-duty status, significantly improving the safety and reliability of fire control room monitoring.

[0166] Please refer to Figure 2 As shown, this application also provides a video anti-reflective personnel on-duty identification device, comprising:

[0167] Acquisition module 201 acquires continuous video images from the camera device;

[0168] The annotation module 202 analyzes the continuous video frame, extracts human image data, and detects and annotates reflection points in the continuous video frame to obtain reflection point annotation data.

[0169] Analysis module 203 performs a comprehensive analysis based on the human image data and the reflection point annotation data in chronological order to generate a record of suspected reflective fragments.

[0170] List module 204, based on the recorded suspected reflective segments, compares the human image contour with the boundary of the reflection point frame by frame to determine the starting area of ​​image misjudgment and generates a list of image misjudgment anchor points;

[0171] Intervention module 205 analyzes the changes in ambient light frequency and the distribution of reflection angles based on the list of image misjudgment anchor points, and generates glare intervention entry data that includes the time window and spatial location range that cause misjudgment.

[0172] Solution module 206 generates a rhythm intervention scheme based on the glare intervention entry data, which is used to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus rebound phase.

[0173] The optimization module 207 adjusts the rhythm and lighting of the video acquisition process according to the rhythm intervention scheme to obtain an optimized video frame sequence.

[0174] The recognition module 208 performs target detection on the optimized video frame sequence and outputs the personnel on duty recognition result.

[0175] This application also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method described above.

[0176] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the above-described method.

[0177] The above embodiments are provided to enable those skilled in the art to understand and apply this application. Those skilled in the art will readily make various modifications to the above embodiments and apply the general principles described herein to other embodiments without inventive effort. Therefore, this application is not limited to the above embodiments, and any improvements and modifications made to this application based on the disclosure thereof should be within the scope of protection of this application.

Claims

1. A method for identifying personnel on duty using video anti-reflective interference, characterized in that, include: Acquire continuous video feeds from a camera device; The continuous video frames are analyzed to extract human image data, and reflection points in the continuous video frames are detected and labeled to obtain reflection point labeling data; Based on the human image data and the reflection point annotation data, a comprehensive analysis is performed in chronological order to generate a record of suspected reflective fragments. Based on the recorded reflective suspected segments, the human image contour and the boundary of the reflection point are compared frame by frame to determine the starting area of ​​image misjudgment and generate a list of image misjudgment anchor points. Based on the list of image misjudgment anchor points, the changes in ambient light frequency and the distribution of reflection angles are analyzed to generate glare intervention entry data that includes the time window and spatial location range that cause misjudgment. Based on the glare intervention entry data, a rhythm intervention scheme is generated to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus rebound phase. According to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted and lighting is intervened to obtain an optimized video frame sequence; Target detection is performed on the optimized video frame sequence, and the on-duty personnel identification result is output.

2. The method according to claim 1, characterized in that, Based on the recorded reflective fragments, the human image contour and the boundary of the reflection point are compared frame by frame to determine the starting area of ​​the image misjudgment, including: Based on the recorded reflective suspected fragments, a frame-by-frame comparison is performed to generate a basic comparison dataset; Based on the aforementioned baseline comparison dataset, the drift path of the human image contour in consecutive frames is calculated; Based on the drift path, the relative motion relationship between the reflection point and the edge of the human image is analyzed to determine the starting area of ​​image misjudgment.

3. The method according to claim 1, characterized in that, Generate a rhythmic intervention scheme for reallocating video capture order, limiting light source switching time points, and correcting the framing rhythm during the camera's focus recovery phase, including: Based on the glare intervention entry data, establish a light intervention benchmark table; Based on the aforementioned illumination intervention reference table, an exposure rhythm instruction is generated; The exposure rhythm command is used to determine the exposure sequence and duration within a stable lighting range to avoid areas of changing lighting, and to reallocate the acquisition order of video frames and limit the timing of light source switching accordingly.

4. The method according to claim 1, characterized in that, Based on the glare intervention entry data, a rhythm intervention scheme is generated to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus recovery phase, including: Generate instructions for performing viewfinder timing corrections during the camera's focus recovery phase; The framing beat correction is used to make the focus recovery process overlap with the stable lighting range.

5. The method according to claim 1, characterized in that, Lighting intervention during video acquisition includes: Based on the reflection area determined by the glare intervention entry data, a reverse-phase dark window projection is performed on the reflection area during the video acquisition phase.

6. The method according to claim 1, characterized in that, According to the rhythm intervention scheme, the video acquisition process is rhythmically adjusted, including: Based on the aforementioned rhythm intervention scheme, the target detection cycle and video acquisition sequence are synchronized.

7. The method according to claim 1, characterized in that, Based on the aforementioned list of misjudged anchor points in the image, the analysis of ambient light frequency variations and reflection angle distribution includes: Based on the list of misjudged anchor points in the image, backtrack the original video frames to extract the image of the reflection point region; Based on the extracted image of the reflection point area, calculate the brightness center and direction of the reflected light, and record the distribution of the reflection angle.

8. A video anti-reflective interference personnel on-duty identification device, characterized in that, include: The acquisition module acquires continuous video footage from the camera device; The annotation module analyzes the continuous video frames, extracts human image data, and detects and annotates reflection points in the continuous video frames to obtain reflection point annotation data. The analysis module performs a comprehensive analysis based on the human image data and the reflection point annotation data in chronological order to generate a record of suspected reflective fragments. The list module compares the human image contour with the boundary of the reflection point frame by frame according to the recorded suspected reflective fragments to determine the starting area of ​​image misjudgment and generate a list of image misjudgment anchor points. The intervention module analyzes the changes in ambient light frequency and the distribution of reflection angles based on the list of image misjudgment anchor points, and generates glare intervention entry data that includes the time window and spatial location range that caused the misjudgment. The solution module generates a rhythm intervention scheme based on the glare intervention entry data, which is used to redistribute the video acquisition order, limit the light source switching time point, and correct the framing rhythm during the camera's focus rebound phase. The optimization module adjusts the rhythm and lighting of the video acquisition process according to the rhythm intervention scheme to obtain an optimized video frame sequence. The recognition module performs target detection on the optimized video frame sequence and outputs the personnel on-duty recognition result.

9. An electronic device, characterized in that, The method includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed in a computer, causes the computer to perform the method described in any one of claims 1 to 7.

Images