Multi-dimensional drowning risk trajectory backtracking visualization system

Abstract

The invention belongs to the technical field of computer vision and behavior recognition, and particularly discloses and provides a multi-dimensional drowning risk trajectory backtracking visualization system, which is characterized in that through time-space correlation analysis of swimmer stroke frequency variation coefficient, heart rate fluctuation trend and water flow data, traditional single data source monitoring limitation is broken through, and early warning mode upgrading is realized; based on spatial relationship detection of dynamic visual range rendering and risk key nodes, responsibility boundaries lacking environmental objective limitation and human intervention are distinguished, and wave height and illumination intensity are introduced to correct transparency of a mask layer in real time, so that the view of a lifeguard is highly consistent with an actual shielding state; a risk weight distributor and a three-dimensional space-time path are utilized to convert chain reactions of action abnormity, water flow environment sudden change and view shielding into a space-time evidence chain capable of being quantitatively traced, multiple data sources are integrated to generate a visual report, and the proof effect of drowning accident liability traceability and the accurate optimization effect of safety design are achieved.

Description

Technical Field

[0001] This invention belongs to the field of computer vision and behavior recognition technology, and relates to a multi-dimensional drowning risk trajectory backtracking visualization system. Background Technology

[0002] Traditional drowning prevention relies primarily on lifeguards' subjective observation and experience, which suffers from response delays and misjudgments due to environmental interference. In recent years, the integration of artificial intelligence, computer vision, and spatiotemporal data analysis technologies has propelled drowning risk prevention towards intelligence, precision, and traceability. Among these advancements, the key to technological breakthroughs in the industry lies in how to achieve real-time retrospective tracking of drowning risk trajectories and determination of responsibility through multi-dimensional data fusion and dynamic visualization technologies.

[0003] Existing safety monitoring primarily relies on video surveillance and sensor technology. Cameras capture swimmer movements, wearable devices monitor physiological indicators, and environmental sensors collect water parameters. While achieving real-time risk warnings and data retention, the following key shortcomings remain: 1. Existing systems treat environmental parameters and swimmer behavior as independent events, lacking quantitative analysis of their spatiotemporal coupling. This may lead to a higher rate of missed detections for complex risks, causing lifeguards to prioritize responses to one independent event while prolonging the average response time for another. 2. Traditional video surveillance only simulates a lifeguard's field of vision with a fixed view frustum, failing to consider the real-time impact of dynamic environmental parameters on the visible range. This may frequently trigger false occlusion alarms due to a lack of dynamic adjustment of the occlusion range, interfering with lifeguards' assessment of actual risks. 3. Existing liability reports only provide timestamps and simple descriptions of risk nodes, lacking cross-validation from multiple data sources and spatiotemporal correlation analysis. This makes it difficult for management to optimize lifeguard deployment or water safety design, leading to recurring similar incidents. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention proposes the following technical solution: a multi-dimensional drowning risk trajectory retrospective visualization system, comprising: a data acquisition module, which acquires in real time the swimmer's stroke frequency variation coefficient, heart rate fluctuation trend, and water flow data, and generates multi-source synchronous time-series data, wherein the water flow data includes water flow velocity and wave height.

[0005] The risk analysis module inputs multi-source synchronous time-series data into the risk weight allocator, dynamically adjusts the weight of the water flow data according to the rhythm of the movement changes, and calculates the spatial occlusion relationship between the swimmer's position and the wave height to output occlusion misjudgment markers, thereby generating a dynamic risk transmission chain.

[0006] The spatiotemporal path marking module obtains the real-time drowning risk value of swimmers and, combined with the dynamic risk transmission chain, constructs a three-dimensional coordinate system for the spatiotemporal path of risk evolution, where the X-axis represents time, the Y-axis represents location, the Z-axis represents drowning risk value, and delineates the boundary of the lifeguard's monitoring blind zone.

[0007] The abnormal feature localization module responds to the drowning event location selection command, traces back the key nodes of risk transmission in the spatiotemporal path of risk evolution, and outputs a group of abnormal skeletal movement features.

[0008] The visual rendering and analysis module dynamically renders and generates a simulation map of the visible range based on lifeguard location data and water flow data, combined with the boundary of the lifeguard's blind spot monitoring area. It also associates abnormal skeletal movement feature groups to generate a responsibility determination.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention breaks through the limitations of traditional single data source monitoring by analyzing the spatiotemporal correlation between the swimmer's stroke frequency variation coefficient, heart rate fluctuation trend and water flow data, and realizes the upgrade of the early warning mode from instant drowning identification to risk transmission chain capture.

[0010] (2) Based on dynamic visible range rendering and spatial relationship detection of risk key nodes, this invention distinguishes the responsibility boundary between objective environmental constraints and lack of human intervention. At the same time, by introducing wave height and light intensity to correct the transparency and range of the mask layer in real time, and combining the light intensity mapping function to dynamically adjust the rendering parameters of the red mask layer, a high degree of consistency between the lifeguard's field of vision and the actual occlusion state is achieved.

[0011] (3) This invention utilizes the synergistic effect of risk weight allocator and three-dimensional spatiotemporal path modeling to transform the chain reaction of abnormal action, sudden change in water flow environment and visual obstruction into a quantifiable and traceable spatiotemporal evidence chain. At the same time, by integrating video surveillance, wearable device and environmental sensor data, a visual report is generated that includes time axis, risk value change curve, obstruction status and cross-validation of multiple data sources, thus achieving the evidentiary effect of tracing responsibility for drowning accidents and the precise optimization effect of safety design. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram of the system module connections of the present invention. Detailed Implementation

[0014] 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.

[0015] Please see Figure 1 As shown, the multi-dimensional drowning risk trajectory retrospective visualization system proposed in this invention includes a data acquisition module, a risk analysis module, a spatiotemporal path marking module, an abnormal feature localization module, and a visualization rendering and analysis module.

[0016] The data acquisition module, risk analysis module, spatiotemporal path marking module, anomaly feature localization module, and visual rendering analysis module are connected in sequence.

[0017] The data acquisition module acquires the swimmer's stroke frequency variation coefficient, heart rate fluctuation trend, and water flow data in real time, generating multi-source synchronous time-series data. The water flow data includes water flow velocity and wave height.

[0018] In a preferred embodiment, the data acquisition module includes: continuously capturing a video stream of a swimmer using a preset ordinary camera, extracting the displacement sequence of skeletal joints from the video stream, and calculating the coefficient of variation of the stroke frequency accordingly. The displacement sequence represents the positional changes of key body parts of the swimmer over time, including but not limited to shoulder joints, elbow joints, knee joints, and head skeletal joints.

[0019] Specifically, the calculation method of the stroke frequency variation coefficient is as follows: Based on the displacement sequence, the stroke action cycle is detected, that is, the start point and end point of each stroke action are identified, and the stroke frequency within each stroke cycle is calculated. The stroke frequency is defined as the number of strokes per unit time. Based on the frequency values ​​of all stroke cycles, the mean μ and standard deviation σ of the frequency are calculated, where μ is obtained by the arithmetic mean of all frequency values, and σ is obtained by the square root of the average of the squares of the deviations of the frequency values ​​from μ. The standard deviation σ is divided by the mean μ and multiplied by 100 to obtain the stroke frequency variation coefficient.

[0020] The raw heart rate signal of swimmers is collected by using a preset low-cost wearable wristband. This signal is a photoplethysmography pulse wave signal. The rate of change of the raw heart rate signal at continuous time points is calculated, that is, the difference in heart rate between adjacent sampling points divided by the time interval.

[0021] If the rate of change is greater than zero, an upward trend indicator is output; if the rate of change is less than zero, a downward trend indicator is output, thus obtaining the heart rate fluctuation trend indicated by the upward or downward trend.

[0022] The time-series data of water flow velocity measured by a preset water flow sensor and the time-series data of wave height measured by a wave height meter together constitute the time-series fluctuation data of water flow data.

[0023] A unified clock chip is used to add timestamps to the time-series fluctuation data of stroke frequency variation coefficient, heart rate fluctuation trend and water flow data to ensure that the time axis of all data points is aligned.

[0024] Multi-source synchronous time-series data is generated by integrating the aligned data.

[0025] The risk analysis module inputs multi-source synchronous time-series data into the risk weight allocator, dynamically adjusts the weight of the water flow data according to the rhythm of the movement changes, and calculates the spatial occlusion relationship between the swimmer's position and the wave height to output occlusion misjudgment markers, thereby generating a dynamic risk transmission chain.

[0026] In a preferred embodiment, the risk analysis module includes: when the coefficient of variation of the stroke frequency exceeds a preset disturbance threshold, activating a water flow data weight multiplier, wherein the water flow data weight multiplier is used to multiply the water flow velocity time series data collected by the water flow sensor by a preset amplification factor to generate weighted water flow data.

[0027] The heart rate fluctuation trend is detected. If the current timestamp output shows an upward trend and the water flow data weight multiplier is active, the wave occlusion analysis unit is triggered.

[0028] The spatial shading relationship between the swimmer's position and wave height is calculated using a wave shading analysis unit. Specifically, the process involves obtaining the real-time wave height measured by a wave altimeter. And obtain the height coordinates of the joints of the swimmer's skull. ,when At that time, calculate the angle between the lifeguard's line of sight and the crest of the wave. The formula is ,in The preset eye level height for lifeguards is obtained by multiplying the lifeguard's height by a preset eye height ratio coefficient, which is 0.92. This ratio coefficient is based on ergonomic standards. The horizontal distance between the swimmer and the lifeguard is calculated using the formula for the planar distance between the center coordinates of the swimmer's skeletal joints and the lifeguard's coordinates. If the value exceeds the preset viewing angle threshold, an occlusion misjudgment mark will be output.

[0029] By integrating the weighted water flow data, occlusion misjudgment markers, and the coefficient of variation of water stroke frequency, a risk transmission link is generated in time stamp order. This risk transmission link is used to record the correlation between abnormal actions, water flow risks, and spatial occlusion relationships at each time point.

[0030] The spatiotemporal path marking module obtains the swimmer's real-time drowning risk value, and constructs a three-dimensional coordinate system risk evolution spatiotemporal path by combining it with the dynamic risk transmission chain. The X-axis represents time, the Y-axis represents location, the Z-axis represents drowning risk value, and the boundary of the lifeguard's monitoring blind zone is delineated.

[0031] In a preferred embodiment, the spatiotemporal path marking module includes: calculating the real-time drowning risk value of a swimmer based on multi-source synchronous time-series data, inputting the risk transmission link into a three-dimensional virtual sandbox, generating a basic three-dimensional path point set, wherein the drowning risk value corresponding to each timestamp is mapped to a Z-axis coordinate value, the timestamp is converted to an X-axis coordinate value, and the center position of the swimmer's skeletal joints is converted to a Y-axis coordinate value.

[0032] Specifically, the drowning risk value is calculated as follows: the heart rate fluctuation trend is converted into a numerical value, such as an upward trend marked as +1, a downward trend marked as -1, and a stable value as 0. The value is then weighted and summed by combining the stroke frequency variation coefficient and the weighted water flow data, i.e.: drowning risk value = w1 × stroke frequency variation coefficient + w2 × heart rate trend converted value + w3 × adjusted water flow velocity + w4 × adjusted wave height, where w1, w2, w3, and w4 are weighting coefficients.

[0033] Linear interpolation is performed on the basic three-dimensional path point set to form a continuous risk evolution spatiotemporal path curve. This curve is visualized in three-dimensional space using color gradients. Specifically, risk threshold ranges are set and associated with three colors: red, orange, and yellow. Red represents a high drowning risk value.

[0034] Read the wave height and the height of the swimmer's head bone joints at each time stamp measured by the wave altimeter. When the wave height exceeds the height of the swimmer's head bone joints, generate a semi-transparent gray column at the Y-axis position of the corresponding time stamp. The column height is equal to the wave height, and the column width covers a preset radius around the swimmer, forming a dynamic environmental shading layer.

[0035] Obtain the preset lifeguard coordinates and calculate the view frustum angle using that point as the vertex. The formula is ,in This indicates the maximum horizontal distance that lifeguards can effectively monitor swimmers in the current environment. It is set as an initial empirical value and then dynamically adjusted based on actual water flow data. For example, in open water... It may be larger; and in a crowded pool, Due to the reduction of obstacles, the area outside the view cone angle boundary in the three-dimensional virtual sand table is marked as the lifeguard monitoring blind zone boundary. The lifeguard monitoring blind zone boundary is represented by a gray semi-transparent area or a dashed box, which intuitively shows the area that lifeguards cannot directly observe as well as the dynamic occlusion area caused by waves.

[0036] The three-dimensional virtual sandbox is the core engine for dynamically simulating and visualizing the spatiotemporal evolution of drowning risk. It is essentially an immersive and interactive three-dimensional digital environment that maps multi-source data into spatial coordinates and drowning risk values ​​to construct a traceable risk transmission model.

[0037] The abnormal feature localization module responds to the drowning event location selection command, traces back the key nodes of risk transmission in the spatiotemporal path of risk evolution, and outputs a group of abnormal skeletal movement features.

[0038] In a preferred embodiment, the response to the drowning event location selection command involves tracing back key nodes of risk transmission along the risk evolution spatiotemporal path and outputting a set of abnormal skeletal movement features. This includes receiving the drowning event location selection command along the risk evolution spatiotemporal path, the command containing information from the time of drowning. and spatial coordinates The spatiotemporal coordinates formed by these points.

[0039] Select the corresponding spatiotemporal coordinates from the location of the drowning incident. The system begins to scan the changes in drowning risk values ​​along the spatiotemporal path curve of risk evolution in reverse along the negative X-axis, with the scan step size being the preset time interval.

[0040] During the scanning process, the current drowning risk value is compared with the preset risk threshold in real time. When the current drowning risk value is detected to be greater than or equal to the preset risk threshold for the first time, the scanning stops and the spatiotemporal coordinate point is recorded as a key node for risk transmission.

[0041] Extract the skeletal movement rhythm data corresponding to the moment when the key node of risk transmission occurs, including the coefficient of variation of the stroke frequency and its duration. At the same time, extract the peak water flow velocity and peak wave height in the water flow data within the same duration.

[0042] The coefficient of variation of stroke frequency, peak water velocity, and peak wave height are combined to output a set of abnormal skeletal movement features.

[0043] In a further preferred embodiment, the step of tracing back key nodes of risk transmission in the spatiotemporal path of risk evolution further includes: during the tracing process, prioritizing the scanning of the heart rate fluctuation trend identifier sequence, and locating the time point at which the first upward trend identifier appears as the starting point of the heart rate increase. .

[0044] The timing of the occurrence of key nodes in risk transmission is recorded as the skeletal movement abnormality point. .

[0045] Compare the temporal relationship between the onset of heart rate rise and the point of skeletal movement abnormality. When the onset of heart rate rise precedes the point of skeletal movement abnormality and the time difference is greater than a preset physiological response threshold, the spatiotemporal coordinates are adjusted. Marked as nodes with latent physiological risk, among which for The spatial coordinates of the swimmer at any given time.

[0046] Extract time-series data of water flow and detect the maximum absolute value of the rate of change of water flow velocity within the time period from the onset of heart rate rise to the point of abnormal skeletal movement. and the maximum absolute value of the rate of change of wave height ,in This indicates the length of time from the onset of heart rate increase to the point of abnormal skeletal movement. This represents the change in water flow velocity over a period of time. This indicates the change in wave height over a given time period.

[0047] When the maximum absolute value of the rate of change of water flow velocity exceeds the threshold for abrupt changes in water flow or the maximum absolute value of the rate of change of wave height exceeds the threshold for abrupt changes in wave height, the extreme point of the water flow data corresponding to the timestamp will be associated with the physiological latent risk node.

[0048] By integrating the onset time of heart rate rise, abnormal points in skeletal movement, and the ratio of the maximum absolute value of the rate of change in water flow data to the mutation threshold, early intervention decision recommendations including suggested intervention time windows are generated.

[0049] Specifically, this invention identifies physiological indicators that precede movement abnormalities by identifying the timing characteristics of these abnormalities. Combined with correlation analysis of sudden changes in water flow data, it reveals latent physiological risks that traditional monitoring methods cannot capture. This provides lifeguards with a basis for decision-making regarding intervention timing before obvious movement abnormalities, thus realizing a shift from a passive response to an active prevention early warning model.

[0050] The visual rendering and analysis module dynamically renders and generates a simulation map of the visible range based on the lifeguard's location data and water flow data, combined with the boundary of the lifeguard's blind spot monitoring area, and generates a responsibility determination based on the abnormal skeletal movement feature group.

[0051] In a preferred embodiment, the step of dynamically rendering a simulation map of the visible range based on lifeguard location data and water flow data, combined with the boundary of the lifeguard's blind spot, includes: acquiring real-time wave height. Preset eye level height for lifeguards Horizontal distance between swimmer and lifeguard Calculate the lower and upper boundary angles visible to the lifeguard, and generate the boundary equation for the cone-shaped region visible to the lifeguard based on these angles; where the lower boundary angle... From the formula Determined, indicating the critical angle at which the lifeguard's line of sight is just not obstructed by the waves; upper boundary angle. From the formula "Confirmed" indicates the maximum angle of repose at which the lifeguard's line of sight covers the swimmer's entire body. This is the preset human height compensation value.

[0052] When the ambient light sensor detects the light intensity parameter At that time, the transparency coefficient is converted to 0-1 through the transparency mapping function. The transparency of the visible cone-shaped region is calculated accordingly, where The preset light intensity threshold, when When the preset light threshold is exceeded, A value of 1 indicates complete opacity; otherwise, the opacity is reduced proportionally.

[0053] In the 3D virtual sandbox, the lifeguard's coordinates are used as vertices, and the visible cone area of ​​the lifeguard is rendered as a semi-transparent cone. The transparency of the cone changes dynamically with the value calculated by the transparency mapping function. The stronger the light, the clearer it is, and the weaker the light, the blurrier it is.

[0054] The boundary of the lifeguard's blind spot is spatially merged with the dynamic environmental masking layer to generate a union area that covers all invisible areas. The union area is then rendered as a red masking layer, and the transparency of the red masking layer is dynamically adjusted in combination with the light intensity parameter.

[0055] Specifically, the range and shape of the merged area and red hood layer are updated in real time as the lifeguard's position, wave height, or swimmer's position changes.

[0056] In a further preferred embodiment, the generation of responsibility attribution for the associated skeletal movement abnormal feature group includes: obtaining the spatiotemporal coordinates of the key nodes of risk transmission and a simulation map of the visible range with a red mask layer, locating the spatiotemporal coordinates of the key nodes of risk transmission in a three-dimensional virtual sandbox, and detecting whether the point is located within the red mask layer area.

[0057] When a critical node for risk transmission is detected to be located in the red masking layer area, the continuous location data within a preset time window before and after the node is analyzed. If all locations are within the red masking layer area, it is determined that the location is continuously within the masking area, triggering the environmental occlusion responsibility report generation unit.

[0058] When a critical node of risk transmission is detected to be outside the red overlay area, the manual intervention missing responsibility report generation unit is triggered.

[0059] The environmental obstruction responsibility report generation unit is used to integrate wave height, light intensity parameters, and lifeguard location data to generate a responsibility report that includes an analysis of environmental obstruction factors.

[0060] The human intervention missing responsibility report generation unit is used to extract the corresponding timestamp skeletal movement abnormal feature group, and combine it with the preset lifeguard reaction time threshold to generate a responsibility report that includes the duration of movement abnormality and lifeguard response delay analysis.

[0061] In a further preferred embodiment, the generation of responsibility attribution for the associated skeletal movement abnormal feature group further includes: when an occlusion misjudgment marker is output, extracting the timestamp corresponding to the marker. Then, the rendering frame with the corresponding timestamp is located in the simulation map of the visible range, and the basic transparency value of the red mask area of ​​the rendering frame is obtained. .

[0062] The new transparency value is calculated using a masking density enhancement algorithm. The calculation formula is: ,in The preset enhancement coefficient, The adjustment range of the abnormal intensity value used to control the rhythm of skeletal movements to the transparency of the red mask layer is usually set to 0.3-0.7, but the specific value needs to be determined through experiments or theoretical calculations. This represents the abnormal intensity value of the coefficient of variation of the stroke frequency. It is obtained by the absolute value of the difference between the coefficient of variation of the stroke frequency and the preset disorder threshold.

[0063] The masking density enhancement algorithm is used to enhance the transparency of the red masking layer to highlight high-risk areas when wave occlusion and abnormal skeletal movements occur simultaneously.

[0064] Update the base transparency value of the red mask area in the rendered frame to the new transparency value. This achieves enhanced masking density.

[0065] Preset time window before and after obtaining the timestamp of the tag Extract the peak timestamps of the coefficient of variation of the stroke frequency from the sequence of stroke frequency variation. Based on this, the time overlap index is calculated. The formula is , is used to quantify the temporal correlation between abnormal peak values ​​of skeletal movements and wave occlusion markers, with a value range of 0 to 1, where 0 indicates no overlap and 1 indicates complete overlap.

[0066] When the time overlap index exceeds the preset overlap threshold, the evidence chain generation unit is triggered.

[0067] The evidence chain generation unit is used to correlate the peak timestamp of the masking concentration, the peak timestamp of the coefficient of variation of the swimming frequency, and the corresponding water flow data to generate a composite evidence chain document containing time correlation analysis. This document provides a scientific, objective, and interactive spatiotemporal analysis tool for tracing responsibility for drowning accidents, significantly improving the accuracy and depth of accident liability determination.

[0068] This invention uses a dynamic coupling enhancement mechanism of masking concentration and abnormal action intensity to intuitively expose the composite risk scenario where wave masking and abnormal action occur simultaneously. The quantitative analysis of time overlap provides objective spatiotemporal correlation evidence for responsibility tracing, breaking through the limitation of the separation of environmental factors and human factors in traditional systems, and realizing the visualization of evidence for multi-factor coupled risk scenarios.

[0069] It should be noted that the formulas described above, through the principle of dimensional consistency and mathematical standardization methods (such as normalization, dimensionless parameter conversion, or unit system unification), can translate physical quantities with different properties into unitless standard values ​​or superimposed parameters of the same dimension. This eliminates the interference of different dimensions on the computational logic, allowing the formulas to retain the original data distribution characteristics while possessing mathematical rationality and adaptability to objective laws. The descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the invention.

Claims

1. A multi-dimensional drowning risk trajectory retrospective visualization system, characterized in that, include: The data acquisition module acquires the swimmer's stroke frequency variation coefficient, heart rate fluctuation trend, and water flow data in real time, and generates multi-source synchronous time-series data, including water flow velocity and wave height. The risk analysis module inputs multi-source synchronous time-series data into the risk weight allocator, dynamically adjusts the weight of the water flow data according to the rhythm of the movement changes, and calculates the spatial occlusion relationship between the swimmer's position and the wave height to output occlusion misjudgment markers, thereby generating a dynamic risk transmission chain. The spatiotemporal path marking module obtains the real-time drowning risk value of swimmers, and constructs a three-dimensional coordinate system risk evolution spatiotemporal path by combining it with the dynamic risk transmission chain. The X-axis represents time, the Y-axis represents location, the Z-axis represents drowning risk value, and the boundary of the lifeguard monitoring blind zone is delineated. The abnormal feature localization module responds to the drowning event location selection command, traces back the key nodes of risk transmission in the spatiotemporal path of risk evolution, and outputs a group of abnormal skeletal movement features. The visual rendering and analysis module dynamically renders and generates a simulation map of the visible range based on lifeguard location data and water flow data, combined with the boundary of the lifeguard's blind spot monitoring area. It also associates abnormal skeletal movement feature groups to generate a responsibility determination.

2. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 1, characterized in that, The data acquisition module includes: The video stream of a swimmer is continuously captured, and the displacement sequence of the skeletal joints is extracted from the video stream to calculate the coefficient of variation of the stroke frequency. By using a pre-set low-cost wearable wristband to collect the swimmer's raw heart rate signal, the rate of change of the raw heart rate signal at continuous time points is calculated. If the rate of change is greater than zero, an upward trend indicator is output; if the rate of change is less than zero, a downward trend indicator is output, thus obtaining the heart rate fluctuation trend indicated by the upward or downward trend. The time-series data of water flow velocity measured by a preset water flow sensor and the time-series data of wave height measured by a wave height meter together constitute the time-series fluctuation data of water flow data. A unified clock chip was used to add timestamps to the time-series fluctuation data of stroke frequency variation coefficient, heart rate fluctuation trend and water flow data to ensure that the time axis of all data points was aligned. Multi-source synchronous time-series data is generated by integrating the aligned data.

3. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 2, characterized in that, The risk analysis module includes the following: When the coefficient of variation of the stroke frequency exceeds the preset disturbance threshold, the water flow data weight multiplier is activated. The heart rate fluctuation trend is detected. If the current timestamp output is an upward trend indicator and the water flow data weight multiplier is active, the wave occlusion analysis unit is triggered. The spatial occlusion relationship between the swimmer's position and the wave height is calculated using the wave occlusion analysis unit to output occlusion misjudgment markers; By integrating the weighted water flow data, occlusion misjudgment markers, and the coefficient of variation of water stroke frequency, a risk transmission link is generated in time stamp order.

4. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 1, characterized in that, The spatiotemporal path marking module includes: Based on multi-source synchronous time-series data, the real-time drowning risk value of swimmers is calculated, and the risk transmission link is input into a three-dimensional virtual sandbox to generate a basic three-dimensional path point set. Linear interpolation is performed on the basic three-dimensional path point set to form a continuous risk evolution spatiotemporal path curve, which is visualized in three-dimensional space using color gradients. Read the wave height and the height of the swimmer's head bone joints at each time stamp measured by the wave altimeter. When the wave height exceeds the height of the swimmer's head bone joints, generate a semi-transparent gray column at the Y-axis position of the corresponding time stamp to form a dynamic environmental occlusion layer. Obtain the preset lifeguard coordinates, calculate the view frustum angle with this point as the vertex, and then mark the area outside the view frustum angle boundary in the 3D virtual sandbox as the lifeguard monitoring blind zone boundary.

5. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 3, characterized in that, The response to the drowning event location selection command traces back to key nodes of risk transmission in the spatiotemporal path of risk evolution, and outputs a set of abnormal skeletal movement features, including: Receive a drowning event location selection instruction in the risk evolution spatiotemporal path. This instruction contains a spatiotemporal coordinate point composed of the drowning time and spatial coordinates. Starting from the location of the drowning incident, select the corresponding spatiotemporal coordinate point and scan the changes in drowning risk value on the spatiotemporal path curve of risk evolution in reverse along the negative X-axis; During the scanning process, the current drowning risk value is compared with the preset risk threshold in real time. When the current drowning risk value is detected to be greater than or equal to the preset risk threshold for the first time, the scanning stops and the spatiotemporal coordinate point is recorded as a key node for risk transmission. Extract the data on the rhythm of skeletal movements corresponding to the moment when the key nodes of risk transmission occur, including the coefficient of variation of the stroke frequency and its duration. At the same time, extract the peak water flow velocity and peak wave height from the water flow data within the same duration. The coefficient of variation of stroke frequency, peak water velocity, and peak wave height are combined to output a set of abnormal skeletal movement features.

6. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 5, characterized in that, The description of tracing key nodes of risk transmission in reverse along the spatiotemporal path of risk evolution also includes: During the reverse tracing process, the heart rate fluctuation trend marker sequence is scanned first, and the time point at which the first upward trend marker appears is identified as the starting point of the heart rate increase. ; The timing of the occurrence of key nodes in risk transmission is recorded as the abnormal point of skeletal movement; Compare the temporal relationship between the onset of heart rate rise and the point of skeletal movement abnormality. When the onset of heart rate rise precedes the point of skeletal movement abnormality and the time difference is greater than a preset physiological response threshold, the spatiotemporal coordinates are adjusted. Marked as nodes with latent physiological risk, among which for The spatial coordinates of the swimmer at any given time; Extract time-series data of water flow and detect the maximum absolute value of the rate of change of water flow velocity and the maximum absolute value of the rate of change of wave height within the time period from the starting point of heart rate rise to the point of abnormal skeletal movement. When the maximum absolute value of the rate of change of water flow velocity exceeds the threshold for abrupt changes in water flow or the maximum absolute value of the rate of change of wave height exceeds the threshold for abrupt changes in wave height, the extreme point of the water flow data corresponding to the timestamp will be associated with the physiological latent risk node. By integrating the onset time of heart rate rise, abnormal points in skeletal movement, and the ratio of the maximum absolute value of the rate of change in water flow data to the mutation threshold, early intervention decision recommendations including suggested intervention time windows are generated.

7. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 4, characterized in that, The method involves dynamically rendering a simulation map of the visible range based on lifeguard location data and water flow data, combined with the boundary of the lifeguard's blind spot. The content includes: By acquiring real-time wave height, preset lifeguard eye level height, and horizontal distance between swimmer and lifeguard, the lower and upper boundary angles of the lifeguard's visibility are calculated, and the boundary equation of the lifeguard's visible cone region is generated accordingly. When the ambient light sensor detects the light intensity parameter, the transparency of the visible cone-shaped area is calculated using the transparency mapping function; In the 3D virtual sandbox, the lifeguard's coordinates are used as vertices, and the visible cone area of ​​the lifeguard is rendered as a semi-transparent cone. The transparency of the cone changes dynamically with the value calculated by the transparency mapping function. The boundary of the lifeguard's blind spot is spatially merged with the dynamic environmental masking layer to generate a union area that covers all invisible areas. The union area is then rendered as a red masking layer, and the transparency of the red masking layer is dynamically adjusted in combination with the light intensity parameter.

8. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 7, characterized in that, The associated skeletal movement abnormality feature group generates a responsibility determination, which includes: Obtain the spatiotemporal coordinates of key risk transmission nodes and a simulated map of the visible range with a red overlay. Locate the spatiotemporal coordinates of key risk transmission nodes in a 3D virtual sandbox and detect whether the point is located within the red overlay area. When a key node of risk transmission is detected to be located in the red masking layer area, the continuous location data within the preset time window before and after the node is analyzed. If all locations are within the red masking layer area, it is determined that the location is continuously within the masking area, triggering the environmental occlusion responsibility report generation unit. When a critical node of risk transmission is detected to be outside the red masking layer area, the manual intervention missing responsibility report generation unit is triggered. The environmental obstruction responsibility report generation unit is used to integrate wave height, light intensity parameters and lifeguard location data to generate a responsibility report that includes environmental obstruction factor analysis. The human intervention missing responsibility report generation unit is used to extract the corresponding timestamp skeletal movement abnormal feature group, and combine it with the preset lifeguard reaction time threshold to generate a responsibility report that includes the duration of movement abnormality and lifeguard response delay analysis.

9. The multi-dimensional drowning risk trajectory retrospective visualization system according to claim 8, characterized in that, The generation of responsibility attribution for the associated skeletal movement abnormality feature group also includes: When an output occlusion misjudgment mark is generated, the timestamp corresponding to the mark is extracted, and then the rendering frame with the corresponding timestamp is located in the visible range simulation map to obtain the basic transparency value of the red masking area of ​​the rendering frame. A new transparency value is calculated using a masking density enhancement algorithm; Update the base transparency value of the red mask area in the rendered frame to a new transparency value to enhance the mask density. Obtain the sequence of paddling frequency variation coefficients within a preset time window before and after the marked timestamp, extract the peak timestamp of the paddling frequency variation coefficients in the sequence, and calculate the time overlap index accordingly. When the time overlap index exceeds the preset overlap threshold, the evidence chain generation unit is triggered. The evidence chain generation unit is used to associate the peak timestamp of the masking concentration, the peak timestamp of the coefficient of variation of the water stroke frequency, and the corresponding water flow data to generate a composite evidence chain document containing time correlation analysis.

Images