Binocular AI infrared safety early warning system for drowning prevention of swimming pool

Abstract

The invention relates to the technical field of drowning prevention early warning, in particular to a binocular AI infrared safety early warning system for drowning prevention of a swimming pool. The swimming posture information of each swimmer is obtained by performing state analysis on the image data, the infrared data and the radar data of the swimmer, and the drowning early warning information is obtained by analyzing the swimming posture information of each swimmer, so that early warning is effectively performed on the drowner, and the fatigue of a lifeguard is reduced; in addition, early warning scheme analysis is carried out on the drowning early warning information to obtain early warning scheme information, rescue execution is carried out according to the early warning scheme information, directional early warning information of a rescue route corresponding to a warning area of a drowning person and a lifeguard in the shortest distance is used as the corresponding early warning scheme information, directional early warning is carried out on the drowning person, and the rescue efficiency is improved. The number of informed people is reduced, the phenomenon that people are prone to consensus psychology under the panic emotion is avoided, and safety is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of anti-drowning warning, and specifically provides a binocular AI infrared safety warning system for preventing drowning in a swimming pool. Background Art

[0002] Drowning is one of the main causes of accidental injury deaths. In a swimming pool environment, due to the special properties of water, once a person loses control, they may face the risk of drowning. Moreover, drowning incidents can bring great psychological trauma to the victims and their families. Therefore, anti-drowning measures through a warning system can effectively reduce the occurrence of drowning accidents, ensure the safety of swimmers, and the warning system can also reduce the workload of lifeguards and reduce the visual fatigue and negligence caused by long-term observation.

[0003] Traditional swimming pool anti-drowning safety warning systems only analyze the movement postures of swimming personnel by collecting video images. However, factors such as the refraction and fluctuation of water bodies can affect the image quality, resulting in a decrease in monitoring accuracy. In addition, when the warning system detects someone drowning, it alerts with a high-decibel alarm. People will panic due to uncertainty about the cause of the alarm. Moreover, in a panicked state, people are prone to herd mentality, and the panic of one person can quickly spread, leading to more people falling into chaos and reducing safety. Summary of the Invention

[0004] The present invention provides a binocular AI infrared safety warning system for preventing drowning in a swimming pool to solve the technical problems mentioned in the above background art.

[0005] In a first aspect of the present invention, a binocular AI infrared safety warning system for preventing drowning in a swimming pool is provided, which includes a data acquisition module, a drowning monitoring and analysis module, a rescue linkage analysis module, a control execution module, and a data management module.

[0006] The data acquisition module includes a binocular camera group installed in the swimming pool venue. The binocular camera group includes a plurality of cameras arranged in the swimming pool venue and has built-in infrared imaging and lidar functions. Fusion image data is obtained through data acquisition by the binocular camera group, and the fusion image data is sent to the drowning monitoring and analysis module. The fusion image data includes the image data of each swimmer collected by the dual camera group, infrared data obtained through the infrared imaging function, and radar data obtained through the lidar function.

[0007] The drowning monitoring and analysis module is used to receive the fused image data and identify the fused image data to obtain the image data, infrared data and radar data of each swimmer, perform status analysis based on the image data, infrared data and radar data of the swimmer to obtain the swimming posture information of each swimmer, and obtain drowning warning information by analyzing the swimming posture information of each swimmer and send the drowning warning information to the rescue linkage analysis module.

[0008] As a further improvement of the present invention, state analysis is performed based on the image data, infrared data and radar data of the swimmer, and the specific analysis method is as follows:

[0009] A quality analysis is performed on the image data of the swimmer to obtain an image quality status. When the image quality status corresponds to a low image quality, infrared data and radar data are obtained, and the swimmer's human thermal signal is obtained based on the infrared data. The human thermal signal of each swimmer is identified to obtain the human thermal signal contour and thermal signal position information of each swimmer. The heat value of each part of the human body is obtained based on the human thermal signal contour at multiple monitoring time points. The heat values ​​corresponding to each part of the human body at multiple monitoring time points are aggregated to obtain a heat change curve of the human body part, and a pre-set heat warning value is obtained. When the heat value corresponding to a part of the human body is less than the heat warning value, the heat value corresponding to the part of the human body and the heat warning value are difference calculated to obtain a heat attenuation value, and the heat attenuation values ​​corresponding to each part of the human body are calculated and summed to obtain a total heat attenuation value.

[0010] The three-dimensional position information of each swimmer is obtained according to the radar data, and the spatial position information of each swimmer is obtained by overlapping the three-dimensional position information of the swimmer and the corresponding thermal signal position information. The stationary time and swimming depth of each swimmer are obtained according to the spatial position information of each swimmer; the pre-set swimming warning depth is obtained, and the part of the swimming depth corresponding to each swimmer that exceeds the swimming warning depth is marked as a risk depth value, and the stationary time of the corresponding swimmer is obtained, and the stationary time is divided into multiple stationary time intervals, each stationary time interval corresponds to a stationary influence coefficient, and the stationary time corresponding to each current swimmer is matched with each stationary time interval to obtain the corresponding stationary influence coefficient.

[0011] The body posture of each swimmer is obtained based on the image data, the corresponding movement amplitude of each swimmer within the unit analysis time period is obtained, and the movement amplitude of each swimmer is identified. When abnormal movement occurs in the movement amplitude, the duration of the abnormal movement is counted and recorded as the abnormal movement duration. Abnormal movements include but are not limited to panic waving of arms, irregular slapping of the water surface, and cessation of leg kicking or rapid and uncoordinated movements.

[0012] The total heat attenuation value, risk depth value, static influence coefficient and abnormal action duration are normalized and their values ​​are taken. Calculate the posture outlier YZ of the swimmer; where RL, YS, FS, and JZ represent the total heat attenuation value, abnormal movement duration, risk depth value, and static influence coefficient respectively, YS' represents the abnormal movement reference duration, and FS' represents the risk depth reference value; G1, G2, and G3 are all preset weight factors, with values of 2.402, 2.245, and 2.589 respectively; Take the posture outlier of each swimmer as the corresponding swimming posture information.

[0013] As a further improvement of the present invention, perform quality analysis on the image data of the swimmer, and the specific analysis method is as follows:

[0014] Obtain the two-dimensional position information, body information, and motion data of each swimmer corresponding to multiple consecutive detection time points according to the image data. Perform position change recognition on the two-dimensional position information of each swimmer corresponding to each monitoring time point to obtain the moving distance of each swimmer corresponding to adjacent detection time points. Calculate the difference between two adjacent moving distances to obtain the moving distance difference. Compare the moving distance difference with a preset distance difference threshold. When the moving distance difference is greater than the distance difference threshold, mark the part of the moving distance difference that exceeds the distance difference threshold as the displacement jump difference.

[0015] Obtain the body posture image of each swimmer according to the body information, and identify the abnormal area of the body image for each swimmer. The body posture image is affected by factors such as light changes, splash occlusion, and water refraction, resulting in blurring, missing, or distortion of the body posture image. Therefore, an abnormal area of the body image will be generated. Calculate the abnormal area corresponding to the abnormal area of the body image of each swimmer, and mark the part where the abnormal area is greater than the preset threshold as the posture outlier.

[0016] Obtain the motion trajectory data of each swimmer corresponding to the analysis time period according to the motion data, and obtain the non-coordination duration corresponding to the trajectory non-coordination state of each swimmer based on the motion trajectory data. The non-coordination state includes but is not limited to jitter, mutation, and abnormal motion speed of the swimmer's motion trajectory; Calculate the ratio of the non-coordination duration to the total duration of the analysis time period to obtain the trajectory outlier.

[0017] Take a reference line segment of a fixed length. Starting from the center point of the reference line segment, draw two straight lines perpendicular to the reference line segment. The two straight lines are respectively located on both sides of the reference line segment, and the lengths of the two straight lines are respectively equal to the displacement jump difference and the attitude anomaly value. Connect the ends of the two straight lines to the two ends of the reference line segment to obtain a quadrilateral. Starting from the center point of the quadrilateral, draw a line segment perpendicular to the quadrilateral. The length of the line segment is equal to the value of the trajectory anomaly. Then, construct a quadrangular pyramid with the quadrilateral and the line segment, calculate the volume of the quadrangular pyramid, and record the value of the volume as the image quality value; when the image quality is greater than a preset threshold, generate a corresponding image quality status as low image quality.

[0018] As a further improvement of the present invention, analyze the swimming posture information of each swimmer. The specific analysis method is as follows:

[0019] Identify the attitude anomaly value of each swimmer from the swimming posture information. When the attitude anomaly value is greater than a preset attitude anomaly reference value, calculate the difference between the attitude anomaly value and the attitude anomaly reference value to obtain the attitude overrun value. Divide the attitude overrun value into multiple attitude overrun value intervals, set an overrun influence coefficient for each attitude overrun value interval, match the current attitude overrun values of each swimmer with each attitude overrun value interval to obtain the corresponding overrun influence coefficient, calculate the product of the attitude overrun value of each swimmer and the corresponding overrun influence coefficient to obtain the drowning risk value. When the drowning risk value is greater than a preset threshold, mark the corresponding swimmer as a drowning person, and use the drowning person as the corresponding drowning warning information.

[0020] The rescue linkage analysis module is used to receive the drowning warning information, analyze the warning plan according to the drowning warning information to obtain the warning plan information, and send the warning plan information to the control execution module.

[0021] As a further improvement of the present invention, analyze the warning plan according to the drowning warning information. The specific analysis method is as follows:

[0022] Obtain the three-dimensional position information corresponding to the drowning person according to the drowning warning information, obtain the pool water surface position point corresponding to the drowning person based on the three-dimensional position information of the drowning person, and obtain the preset warning radius. Based on the warning radius and the pool water surface position point corresponding to the drowning person, obtain the warning area and conduct directional warning for the warning area; obtain the position information of each lifeguard based on data management, match the position information of each lifeguard with the warning area corresponding to the drowning person to obtain the lifeguard with the shortest distance, and mark the established route between the lifeguard with the shortest distance and the warning area as the rescue route, and conduct directional warning for the rescue route. Use the directional warning information of the warning area of the drowning person and the rescue route corresponding to the lifeguard with the shortest distance as the corresponding warning plan information; the way of directional warning includes using the light curtain and audio thrown in a directional manner to remind the other swimmers in the rescue route and the warning area to avoid, and it is convenient for the lifeguard to quickly reach near the drowning person through the preset rescue route for rescue.

[0023] The control execution module is used to receive the warning plan information and perform rescue execution according to the warning plan information, send rescue alarm information to the lifeguard with the shortest distance corresponding to the drowning person, and perform sound and light projection reminder according to the directional warning corresponding to the warning area and the rescue route. The sound and light projection reminder is in a prominent reminder color and includes a low-volume voice broadcast reminder; the rescue alarm information includes silent vibration alarm through intelligent wearable devices to avoid large-scale panic.

[0024] The data management module is used to store the image data, infrared data and radar data of each swimmer, and store the drowning warning information and the warning plan information.

[0025] In the technical solution provided by the present invention, compared with the prior art, the beneficial effects are as follows:

[0026] 1. In the present invention, the image data, infrared data and radar data of each swimmer are obtained by identifying the fused image data, the swimming posture information of each swimmer is obtained through state analysis based on the image data, infrared data and radar data of the swimmer, and the drowning warning information is obtained by analyzing the swimming posture information of each swimmer, effectively warning the drowning person and reducing the fatigue of the lifeguard.

[0027] 2. In the present invention, the warning plan information is obtained by receiving the drowning warning information and performing warning plan analysis according to the drowning warning information, and the rescue execution is performed according to the warning plan information. The directional warning information of the warning area of the drowning person and the rescue route corresponding to the lifeguard with the shortest distance is used as the corresponding warning plan information to conduct directional warning for the drowning person, reducing the informed people and avoiding the herd mentality that is prone to occur among the crowd under panic, increasing safety. Description of the Drawings

[0028] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the description of the embodiments. The following drawings are not deliberately drawn to scale according to the actual size, and the focus is on showing the gist of the present application.

[0029] Figure 1 It is a principle block diagram of a binocular AI infrared safety warning system for preventing drowning in a swimming pool according to the present invention;

[0030] Figure 2 It is a schematic diagram of the curve of the heat change of the human body parts of a binocular AI infrared safety warning system for preventing drowning in a swimming pool according to the present invention. Specific embodiments

[0031] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0032] For ease of understanding, the following describes the specific process of the embodiments of the present invention. Please refer to Figure 1-2 In an embodiment of the present invention, an embodiment of a binocular AI infrared safety warning system for preventing drowning in a swimming pool includes:

[0033] The data acquisition module includes a binocular camera group installed in the swimming pool site. The binocular camera group includes a plurality of cameras arranged in the swimming pool site and has built-in infrared imaging and lidar functions. The fused image data is obtained through data acquisition by the binocular camera group, and the fused image data is sent to the drowning monitoring and analysis module; the fused image data includes the image data of each swimmer collected by the binocular camera group, and the infrared data obtained through the infrared imaging function, and the radar data obtained through the lidar function.

[0034] The drowning monitoring and analysis module receives the fused image data and identifies the fused image data to obtain the image data, infrared data and radar data of each swimmer, analyzes the state according to the image data, infrared data and radar data of the swimmer to obtain the swimming posture information of each swimmer, and analyzes the swimming posture information of each swimmer to obtain the drowning warning information and sends the drowning warning information to the rescue linkage analysis module.

[0035] The state analysis is performed according to the image data, infrared data and radar data of the swimmer. The specific analysis method is as follows:

[0036] Perform quality analysis on the image data of the swimmers to obtain the image quality status. When the image quality status corresponds to low image quality, obtain the infrared data and radar data. Obtain the human body heat signal of the swimmers based on the infrared data. Identify the human body heat signal contours and heat signal position information of each swimmer by analyzing the human body heat signals of each swimmer. Obtain the heat values of each part of the human body based on the human body heat signal contours at multiple monitoring time points. Aggregate the heat values corresponding to multiple monitoring time points for each part of the human body to obtain the heat change curve of the human body parts. Obtain the preset heat warning value. When the heat value corresponding to a part of the human body is less than the heat warning value, calculate the difference between the heat value corresponding to the human body part and the heat warning value to obtain the heat attenuation value. Calculate the sum of the heat attenuation values corresponding to each part of the human body to obtain the total heat attenuation value.

[0037] Obtain the three-dimensional position information of each swimmer based on the radar data. Overlay the three-dimensional position information of the swimmers and the corresponding heat signal position information to obtain the spatial position information of each swimmer. Obtain the stationary time and swimming depth of each swimmer based on the spatial position information of each swimmer. Obtain the preset swimming warning depth. Mark the part of the swimming depth of each swimmer that exceeds the swimming warning depth as the risk depth value, and obtain the stationary time of the corresponding swimmer. Divide the stationary time into multiple stationary time intervals, and each stationary time interval corresponds to a stationary influence coefficient. Match the current stationary time of each swimmer with each stationary time interval to obtain the corresponding stationary influence coefficient.

[0038] Obtain the body postures of each swimmer based on the image data. Obtain the movement amplitude of each swimmer corresponding to the unit analysis time period. Identify the movement amplitude of each swimmer. When an abnormal movement occurs in the movement amplitude, count the duration of the abnormal movement and record it as the abnormal movement duration. Abnormal movements include, but are not limited to, the arms flailing wildly, irregularly slapping the water surface, and the legs stopping kicking or becoming rapid and uncoordinated.

[0039] Normalize the total heat attenuation value, risk depth value, stationary influence coefficient, and abnormal movement duration and take their numerical values. Use the preset model to calculate the posture abnormality value YZ of the swimmer; where RL, YS, FS, and JZ represent the total heat attenuation value, abnormal movement duration, risk depth value, and stationary influence coefficient respectively, YS' represents the abnormal movement reference duration, and FS' represents the risk depth reference value; G1, G2, and G3 are all preset weight factors, and their magnitudes are customarily set, such as taking the values 2.402, 2.245, and 2.589 respectively, and the model and weight factors are pre-stored in the drowning monitoring and analysis module; regard the posture abnormality value of each swimmer as the corresponding swimming posture information.

[0040] Perform quality analysis on the image data of swimmers, and the specific analysis method is as follows:

[0041] Obtain the two-dimensional position information, body information, and motion data of each swimmer corresponding to multiple consecutive detection time points according to the image data. Perform position change recognition on the two-dimensional position information of each swimmer corresponding to each monitoring time point to obtain the moving distance of each swimmer corresponding to adjacent detection time points. Calculate the difference between two adjacent moving distances to obtain the moving distance difference. Compare the moving distance difference with a pre-set distance difference threshold. When the moving distance difference is greater than the distance difference threshold, mark the part of the moving distance difference that exceeds the distance difference threshold as the displacement jump difference.

[0042] Obtain the body posture images of each swimmer according to the body information. Identify the abnormal regions of the body posture images of each swimmer. The body posture images are blurred, missing, or distorted due to factors such as light changes, splash occlusion, and water refraction, so abnormal regions of the body images will be generated. Calculate the area of the abnormal regions corresponding to the body image abnormal regions of each swimmer, and mark the part where the area of the abnormal region is greater than the preset threshold as the posture abnormal value.

[0043] Obtain the motion trajectory data of each swimmer corresponding to the analysis time period according to the motion data. Obtain the non-coordination duration corresponding to the trajectory non-coordination state of each swimmer based on the motion trajectory data. The non-coordination state includes but is not limited to the jitter, mutation, and abnormal motion speed of the swimmer's motion trajectory. Calculate the ratio of the non-coordination duration to the total duration of the analysis time period to obtain the trajectory abnormal value.

[0044] Take a reference line segment of a fixed length. Starting from the center point of the reference line segment, draw two straight lines perpendicular to the reference line segment. The two straight lines are located on both sides of the reference line segment respectively, and the lengths of the two straight lines are equal to the displacement jump difference and the posture abnormal value respectively. Connect the ends of the two straight lines to the two ends of the reference line segment to obtain a quadrilateral. Starting from the center point of the quadrilateral, draw a line segment perpendicular to the quadrilateral. The length of the line segment is equal to the value of the trajectory abnormal value. Then construct a quadrangular pyramid with the quadrilateral and the line segment, calculate the volume of the quadrangular pyramid, and record the value of the volume as the image quality value. When the image quality is greater than the preset threshold, generate the corresponding image quality state as low image quality.

[0045] Analyze the swimming posture information of each swimmer, and the specific analysis method is as follows:

[0046] Identify the abnormal posture values of each swimmer from the swimming posture information. When the abnormal posture value is greater than the preset abnormal posture reference value, calculate the difference between the abnormal posture value and the abnormal posture reference value to obtain the posture overrun value. Divide the posture overrun value into multiple posture overrun value intervals, set an overrun impact coefficient for each posture overrun value interval, match the current posture overrun values of each swimmer with each posture overrun value interval to obtain the corresponding overrun impact coefficient, calculate the product of the posture overrun value of each swimmer and the corresponding overrun impact coefficient to obtain the drowning risk value. When the drowning risk value is greater than the preset threshold, mark the corresponding swimmer as a drowning victim and use the drowning victim as the corresponding drowning warning information.

[0047] The rescue linkage analysis module receives the drowning warning information, analyzes the warning plan based on the drowning warning information to obtain the warning plan information, and sends the warning plan information to the control execution module.

[0048] Analyze the warning plan based on the drowning warning information. The specific analysis method is as follows:

[0049] Obtain the three-dimensional position information corresponding to the drowning victim based on the drowning warning information, obtain the pool water surface position point corresponding to the drowning victim according to the three-dimensional position information of the drowning victim, and obtain the preset warning radius. Based on the warning radius and the pool water surface position point corresponding to the drowning victim, obtain the warning area and conduct directional warning for the warning area; obtain the position information of each lifeguard based on data management, match the position information of each lifeguard with the warning area corresponding to the drowning victim to obtain the lifeguard with the shortest distance, and mark the established route of the lifeguard with the shortest distance and the warning area as the rescue route, and conduct directional warning for the rescue route. Use the directional warning information of the warning area of the drowning victim and the rescue route corresponding to the lifeguard with the shortest distance as the corresponding warning plan information; the method of directional warning includes using a light curtain and audio thrown in a specific direction to remind the other swimmers in the rescue route and the warning area to avoid, and it is convenient for the lifeguard to quickly reach near the drowning victim through the preset rescue route for rescue.

[0050] The control execution module receives the warning plan information, performs rescue operations according to the warning plan information, sends rescue alarm information to the lifeguard with the shortest distance corresponding to the drowning victim, and conducts sound and light projection reminders according to the directional warnings corresponding to the warning area and the rescue route. The sound and light projection reminder is in a prominent reminder color and includes a low-volume voice broadcast reminder; the rescue alarm information includes silent vibration alarm through intelligent wearable devices to avoid large-scale panic.

[0051] The data management module stores the image data, infrared data, and radar data of each swimmer, as well as the drowning warning information and the warning plan information.

[0052] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A binocular AI infrared safety warning system for preventing drowning in a swimming pool, including a data management module, characterized in that, It further includes: A data acquisition module, which is used to acquire fused image data and send the fused image data to the drowning monitoring and analysis module; A drowning monitoring and analysis module, which is used to receive the fused image data, identify the image data, infrared data, and radar data of each swimmer from the fused image data, perform status analysis based on the image data, infrared data, and radar data of the swimmer to obtain the swimming posture information of each swimmer, and analyze the swimming posture information of each swimmer to obtain a drowning warning message and send the drowning warning message to the rescue linkage analysis module; A rescue linkage analysis module, which is used to receive the drowning warning message, perform warning plan analysis based on the drowning warning message to obtain warning plan information, and send the warning plan information to the control execution module; A control execution module, which is used to receive the warning plan information and perform rescue execution based on the warning plan information.

2. The binocular AI infrared safety warning system for preventing drowning in a swimming pool according to claim 1, characterized in that, For acquiring the fused image data and sending the fused image data to the drowning monitoring and analysis module, specifically: a binocular camera group installed in the pool venue, the binocular camera group includes multiple cameras arranged in the pool venue, and is built-in with infrared imaging and lidar functions, and the fused image data is acquired through the binocular camera group; the fused image data includes the image data of each swimmer acquired by the dual camera group, and the infrared data obtained through the infrared imaging function, and the radar data obtained through the lidar function.

3. An AI infrared safety warning system with binoculars for preventing drowning in a swimming pool according to claim 1, characterized in that, The specific analysis method for performing status analysis based on the image data, infrared data, and radar data of the swimmer is as follows: Perform quality analysis on the image data of the swimmer to obtain the image quality status. When the image quality status corresponds to low image quality, obtain the infrared data and radar data, obtain the human body heat signal of the swimmer according to the infrared data, identify the human body heat signal contour and heat signal position information of each swimmer through the human body heat signal of each swimmer, obtain the heat value of each part of the human body according to the human body heat signal contour at multiple monitoring time points, aggregate the heat values corresponding to multiple monitoring time points of each part of the human body to obtain the human body part heat change curve, obtain the preset heat warning value, when the heat value corresponding to the part of the human body is less than the heat warning value, calculate the difference between the heat value corresponding to the part of the human body and the heat warning value to obtain the heat attenuation value, and calculate and sum the heat attenuation values corresponding to each part of the human body to obtain the total heat attenuation value; Obtain the three-dimensional position information of each swimmer according to the radar data, overlap the three-dimensional position information of the swimmer and the corresponding heat signal position information to obtain the spatial position information of each swimmer, and obtain the static time and swimming depth of each swimmer according to the spatial position information of each swimmer; obtain the preset swimming warning depth, mark the part where the swimming depth of each swimmer exceeds the swimming warning depth as the risk depth value, and obtain the static time of the corresponding swimmer, divide the static time into multiple static time intervals, each static time interval corresponds to a static influence coefficient, and match the current static time corresponding to each swimmer with each static time interval to obtain the corresponding static influence coefficient; Obtain the body postures of each swimmer based on the image data, obtain the movement amplitudes corresponding to each swimmer within a unit analysis time period, identify the movement amplitudes of each swimmer, and when an abnormal movement occurs in the movement amplitude, count the duration of the abnormal movement and record it as the abnormal movement duration; comprehensively calculate the total heat attenuation value, risk depth value, static influence coefficient, and abnormal movement duration to obtain the posture abnormality value of the swimmer; use the posture abnormality value of each swimmer as the corresponding swimming posture information.

4. The binocular AI infrared safety warning system for preventing drowning in a swimming pool according to claim 3, characterized in that, The comprehensive calculation of the total heat attenuation value, risk depth value, static influence coefficient, and abnormal operation duration is specifically as follows: The total heat attenuation value, risk depth value, static influence coefficient, and abnormal operation duration are normalized and their numerical values are taken, and a preset model is used to calculate the posture abnormality value YZ of the swimmer; where RL, YS, FS, and JZ respectively represent the total heat attenuation value, abnormal operation duration, risk depth value, and static influence coefficient, YS' represents the abnormal operation reference duration, and FS' represents the risk depth reference value; G1, G2, and G3 are all preset weight factors.

5. The binocular AI infrared safety warning system for preventing drowning in a swimming pool according to claim 1, characterized in that, The quality analysis of the image data of the swimmers is carried out in the following specific analysis method: Based on the image data, obtain the two-dimensional position information, body information, and motion data corresponding to each swimmer at multiple consecutive detection time points. Perform position change identification on the two-dimensional position information of each swimmer at each monitoring time point to obtain the moving distance corresponding to each swimmer at adjacent detection time points. Calculate the difference between two adjacent moving distances to obtain the moving distance difference. Compare the moving distance difference with a pre-set distance difference threshold. When the moving distance difference is greater than the distance difference threshold, mark the part of the moving distance difference that exceeds the distance difference threshold as the displacement jump difference; Obtain the body posture images of each swimmer based on the body information, identify the abnormal regions of the body posture images of each swimmer, calculate the abnormal region area corresponding to the abnormal regions of the body images of each swimmer, and mark the part where the abnormal region area is greater than the preset threshold as the posture abnormality value; Based on the motion data, obtain the motion trajectory data corresponding to each swimmer in the analysis time period, obtain the non-coordination duration corresponding to the trajectory non-coordination state of each swimmer based on the motion trajectory data, and calculate the ratio of the non-coordination duration to the total duration of the analysis time period to obtain the trajectory abnormality value; Comprehensively analyze the displacement jump difference, posture abnormality value, and trajectory abnormality value to obtain the image quality value. When the image quality is greater than the preset threshold, generate the corresponding image quality status as low image quality.

6. The binocular AI infrared safety warning system for preventing drowning in a swimming pool according to claim 5, characterized in that, The specific analysis method for comprehensively analyzing the displacement jump difference, posture abnormality value, and trajectory abnormality value is as follows: Take a reference line segment of a fixed length. Starting from the center point of the reference line segment, draw two straight lines perpendicular to the reference line segment. The two straight lines are located on both sides of the reference line segment respectively, and the lengths of the two straight lines are equal to the displacement jump difference and the posture abnormality value respectively. Connect the ends of the two straight lines to the two ends of the reference line segment to obtain a quadrilateral. Starting from the center point of the quadrilateral, draw a line segment perpendicular to the quadrilateral. The length of the line segment is equal to the value of the trajectory abnormality value. Then, construct a quadrangular pyramid with the quadrilateral and the line segment, calculate the volume of the quadrangular pyramid, and record the value of the volume as the image quality value.

7. An AI binocular infrared safety warning system for preventing drowning in a swimming pool according to claim 1, characterized in that, The specific analysis method for analyzing the swimming posture information of each swimmer is as follows: Identify the abnormal posture values of each swimmer by recognizing the swimming posture information. When the abnormal posture value is greater than the preset abnormal posture reference value, calculate the difference between the abnormal posture value and the abnormal posture reference value to obtain the posture overrun value. Divide the posture overrun value into multiple posture overrun value intervals, set an overrun influence coefficient for each posture overrun value interval, match the current posture overrun values of each swimmer with each posture overrun value interval to obtain the corresponding overrun influence coefficient, calculate the product of the posture overrun value of each swimmer and the corresponding overrun influence coefficient to obtain the drowning risk value. When the drowning risk value is greater than the preset threshold, mark the corresponding swimmer as a drowning victim and use the drowning victim as the corresponding drowning warning information.

8. The binocular AI infrared safety warning system for preventing drowning in a swimming pool according to claim 1, characterized in that, The specific analysis method for analyzing the warning plan according to the drowning warning information is as follows: Obtain the three-dimensional position information corresponding to the drowning victim according to the drowning warning information, obtain the pool water surface position point corresponding to the drowning victim according to the three-dimensional position information of the drowning victim, and obtain the preset warning radius. Based on the warning radius and the pool water surface position point corresponding to the drowning victim, obtain the warning area and conduct directional warning for the warning area; Based on data management, obtain the position information of each lifeguard, match the position information of each lifeguard with the warning area corresponding to the drowning victim to obtain the lifeguard with the shortest distance, and mark the established route between the lifeguard with the shortest distance and the warning area as the rescue route, and conduct directional warning for the rescue route. Use the directional warning information of the warning area of the drowning victim and the rescue route corresponding to the lifeguard with the shortest distance as the corresponding warning plan information.

9. An AI binocular infrared safety warning system for preventing drowning in a swimming pool according to claim 1, characterized in that, The specific execution method for receiving the warning plan information and performing rescue according to the warning plan information is as follows: Send a rescue alarm message to the lifeguard with the shortest distance corresponding to the drowning victim, and conduct sound and light projection reminder according to the directional warning corresponding to the warning area and the rescue route. The sound and light projection reminder is in a prominent reminder color and includes a low-volume voice broadcast reminder.

Images