Swimming pool environment drowning risk identification and control system based on vtn model

Abstract

The application discloses a pool environment drowning risk identification management and control system based on a VTN model, belongs to the technical field of pool environment drowning identification, and comprises a platform end and a user end. The platform end comprises a pool module and a model establishment module. The user end comprises a pool information module, a monitoring module and a risk identification module. The pool module is used for pool analysis and the generation of a regional distribution map. The model establishment module is used for establishing a VTN model based on the regional distribution map, adjusting the regional distribution map according to the VTN model, and obtaining a benchmark distribution map. The pool information module is used for collecting pool information, sending the pool information to the pool module of the platform end, collecting influence material data according to influence identification characteristics, and displaying the benchmark distribution map to users. The monitoring module is used for real-time monitoring of a pool and the acquisition of monitoring data. The risk identification module is used for real-time analysis of the monitoring data based on the VTN model and the acquisition of corresponding risk identification results.

Description

Technical Field

[0001] This invention belongs to the field of swimming pool environment drowning identification technology, specifically a swimming pool environment drowning risk identification and control system based on the VTN model. Background Technology

[0002] As people's interest in swimming continues to grow, indoor swimming pools have become a popular place for most people to relax after work. However, while people enjoy swimming, drowning has become a significant problem. Currently, drowning prevention and rescue in swimming pools mainly rely on lifeguards' on-site patrols and rescue efforts. But this method has many limitations. On the one hand, lifeguards' attention may be affected by fatigue, distraction, and other factors, leading to blind spots in monitoring. On the other hand, in large or crowded pools, lifeguards find it difficult to monitor the status of all swimmers simultaneously, especially when swimmers are obstructed by other people, making it easier to miss them.

[0003] With the continuous development of artificial intelligence and computer vision technologies, drowning prevention and detection algorithms based on visual analysis have gradually become a research hotspot. These algorithms capture images of the water in real time using cameras, and then analyze and process the images using deep learning algorithms to detect the behavior and status of people in real time. When a potential drowning risk is detected, the system immediately issues an alarm, notifying on-site lifeguards to take emergency measures. However, existing target detection technologies suffer from low efficiency, high false alarm rates, and insufficient real-time performance, making it difficult to meet the needs of precise prevention and control.

[0004] Based on this, in order to solve the problem of accurate identification of drowning in swimming pool environments, this invention provides a swimming pool environment drowning risk identification and control system based on the VTN model. Summary of the Invention

[0005] To address the problems of the above solutions, this invention provides a swimming pool environment drowning risk identification and control system based on the VTN model.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A swimming pool environment drowning risk identification and control system based on the VTN model includes a platform and a user terminal; communication connections exist between the platform and the user terminals of each user.

[0008] The platform includes a pool module and a model building module;

[0009] The pool module is used to perform pool analysis, obtain pool information sent by the corresponding user terminal, and obtain impact analysis data of the preset influencing factors based on the pool information;

[0010] Based on the impact analysis data, determine whether the swimming pool is affected by the influencing factors, and obtain the impact judgment result of the influencing factors, which includes whether there is an impact or no impact; mark the influencing factors with the impact judgment result of having an impact as target factors;

[0011] The pool area is divided into zones based on the corresponding target factors to obtain a regional distribution map of the target factors.

[0012] Furthermore, based on the impact analysis data, it is determined whether the swimming pool is affected by the corresponding influencing factors, including:

[0013] The platform provides an impact identification library, which stores the impact identification features corresponding to each influencing factor.

[0014] An impact calibration model is established based on the impact identification library. The expression for the impact calibration model is as follows:

[0015] ;

[0016] In the formula: s i This represents the impact analysis data of the corresponding influencing factors, where i represents the corresponding influencing factor, i = 1, 2, ..., n, and n is the number of influencing factors; the output data is the impact calibration value PH(s). i This affects the calibration value to be 1 or 0;

[0017] The influence calibration value of the influencing factors is obtained by analyzing the influence analysis data of the corresponding influencing factors through the influence calibration model.

[0018] When the influence calibration value is 1, the influence judgment result of the influencing factor is that it has an influence;

[0019] When the influence calibration value is 0, the influence of the influencing factor is judged as having no influence.

[0020] Furthermore, the pool area is divided into zones based on target factors, including:

[0021] Step SA1: Generate a pool area map of the pool area, obtain the influence identification features of the target factors, send the influence identification features of the target factors to the user terminal of the corresponding user, and receive the influence material data sent by the corresponding user terminal; mark the feature identification data of the influence factors that have an impact on the corresponding location in the pool area map according to the influence material data;

[0022] Step SA2: Merge adjacent locations with identical feature recognition data in the pool area map to obtain several unit areas;

[0023] Step SA3: Evaluate whether adjacent unit regions meet the merging requirements, merge adjacent unit regions that meet the merging requirements, and obtain new unit regions;

[0024] Step SA4: Repeat step SA3 until there are no adjacent cell regions that meet the merging requirements, then mark the cell region as a segmented region; mark the current pool area map as a regional distribution map.

[0025] Further, in step SA3, the evaluation of whether adjacent unit regions meet the merging requirements includes:

[0026] Identify the feature recognition data corresponding to the unit region, and determine the influence value of the unit region based on the feature recognition data;

[0027] Calculate the absolute value of the difference between the corresponding influence values ​​of adjacent unit regions and mark it as the influence difference;

[0028] The impact difference is used to determine whether the merging requirements between adjacent unit areas are met.

[0029] Furthermore, when there are multiple different feature recognition data for the same location or the same unit area in the pool area map, the feature recognition data that has the greatest impact on drowning identification in the VTN model is retained.

[0030] The model building module is used to build a VTN model based on the regional distribution map, adjust the regional distribution map according to the VTN model to obtain the corresponding baseline distribution map, send the regional distribution map to the pool information module of the corresponding user terminal, and deploy the VTN model in the risk identification module of the corresponding user terminal.

[0031] Furthermore, the regional distribution map is adjusted according to the VTN model, including:

[0032] Obtain feature recognition data for each location in the regional distribution map, and generate corresponding verification data based on the feature recognition data;

[0033] The VTN model was used to analyze the validation data to obtain the influence value of the VTN model at the corresponding locations.

[0034] The segmented regions in the regional distribution map are adjusted based on the influence values ​​of each location, and the influence values ​​corresponding to each segmented region are added. The adjusted regional distribution map is then marked as the baseline distribution map.

[0035] The user terminal includes a pool information module, a monitoring module, and a risk identification module;

[0036] The pool information module is used to collect pool information, send the pool information to the pool module on the platform, and collect influence material data based on the influence identification features of the received target factors. The influence material data consists of feature identification data. The influence material data is sent to the pool module on the platform. The received baseline distribution map is displayed to the user.

[0037] The monitoring module is used to monitor the swimming pool in real time, obtain the corresponding monitoring data, and send the monitoring data to the risk identification module.

[0038] The risk identification module is used to perform real-time analysis of monitoring data based on the VTN model, obtain corresponding risk identification results, and perform corresponding processing based on the risk identification results.

[0039] The area display module is used to identify and analyze the risk status of the pool area, obtain a risk situation map, and display the risk situation map to the pool management personnel.

[0040] Furthermore, a risk identification status analysis of the swimming pool area is conducted, including:

[0041] Acquire monitoring data, identify the feature data that affects drowning identification in the VTN model based on the monitoring data, and determine the influence value of the VTN model under the corresponding feature data.

[0042] Based on the influence values ​​of each location, the pool area map is merged to obtain several segmented areas, and the corresponding influence values ​​are marked for each segmented area; the pool area map is then marked as a risk situation map.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] Through the cooperation between the various modules of this invention, accurate identification of swimming pool drowning risks can be achieved.

[0045] The VTN model, through its self-attention mechanism and Transformer architecture, can capture global spatiotemporal dependencies in videos, effectively identifying subtle changes in drowning behavior (such as slow sinking or abnormal stillness). By clearly defining target factors and regional distribution maps, it facilitates targeted VTN model building, improving model building efficiency and recognition accuracy. Presenting the received baseline distribution map to users provides an initial, intuitive understanding of the risk situation in the pool area. Dynamically displaying risk situation maps to pool-related personnel enables users and managers to more effectively manage pool risks, strengthen management of high-impact areas, improve drowning recognition accuracy, and compensate for potential intelligent recognition errors. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0047] Figure 1 This is a block diagram illustrating the principle of the present invention;

[0048] Figure 2 This is an example diagram illustrating the technical route of the VTN model of this invention. Detailed Implementation

[0049] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0050] like Figures 1 to 2 As shown, the swimming pool environment drowning risk identification and control system based on the VTN model includes a platform and a user terminal.

[0051] The user terminal is used by users for pool management and other purposes; the platform terminal is used by the platform provider, and the platform terminal communicates with the user terminals of each user.

[0052] The platform includes a pool module and a model building module;

[0053] The pool module is used to perform pool analysis and obtain pool information sent by the corresponding user terminal. The pool information includes relevant information such as pool size, lighting layout, and monitoring equipment layout. The monitoring equipment layout information refers to information that meets the monitoring requirements, such as the camera being high-definition (1080P and above) and covering key areas of the pool (such as the deep water area and children's area). The specific information is determined according to the monitoring requirements. If the monitoring requirements are not met, the monitoring equipment layout information is obtained after the monitoring equipment is adjusted.

[0054] The platform determines influencing factors affecting drowning detection based on historical data from the VTN model, such as glare, obstruction, and lighting changes. Impact analysis data is then obtained based on these factors and pool information. For example, for glare, the impact analysis data consists of data related to pool glare, typically historical video surveillance data from monitoring equipment, data on the causes of glare, etc., covering various scenarios. Therefore, monitoring data from multiple days is generally selected. This monitoring data can also provide analytical material for factors like obstruction and lighting. In other words, the impact analysis data is multifaceted; for example, lighting changes include not only monitoring data but also other factors within the building that affect pool brightness (whether...). The impact analysis data includes factors such as external light transmission and lighting equipment information. Based on this data, it is determined whether the pool is affected by a particular factor. For example, regarding lighting, if the pool is indoors and not affected or minimally affected by external light, and the lighting equipment provides stable illumination, then it will not be affected by changes in light intensity. Generally, there is a reflective effect, but if redundant setups with multiple monitoring devices are used to overcome this, and multi-video stitching technology is used to obtain a non-reflective monitoring video of the pool, then it will not be affected by reflection. The specific determination depends on the actual situation of the pool. The impact analysis data is used to determine the impact of each factor, including whether it has an impact or not. Factors with an impact determination of "having an impact" are marked as target factors.

[0055] The pool area is divided into zones based on the corresponding target factors to obtain a regional distribution map of the target factors.

[0056] By clearly defining the target factors and regional distribution maps, it is easier to build a targeted VTN model, thereby improving the efficiency of VTN model building and identification accuracy.

[0057] In one embodiment, the determination of whether the swimming pool is affected by the corresponding influencing factors can be based on the impact analysis data. This can be done using existing judgment methods, such as the common use of machine learning and deep learning algorithms to build intelligent models for intelligent judgment.

[0058] In one embodiment, determining whether the swimming pool is affected by a corresponding influencing factor based on impact analysis data includes:

[0059] The platform collects historical drowning identification data of the VTN model under various influencing factors, determines the influence identification features of the corresponding influencing factors on drowning identification of the VTN model based on the historical drowning identification data, classifies and stores each influence identification feature according to the influencing factor, and marks the stored database as the influence identification library.

[0060] An impact calibration model is established based on the impact identification database. This model is used to identify features of the impact analysis data of corresponding influencing factors based on the database. It determines whether the data is affected by the influencing factors based on whether the corresponding influencing factors in the database possess their characteristic impact features. Furthermore, it determines whether the impact analysis data meets the impact criteria, which indicates that the data is affected by the influencing factors. The expression for the impact calibration model is:

[0061] ;

[0062] In the formula: s i The input data represents the impact analysis data of the corresponding influencing factors, i = 1, 2, ..., n, where n is the number of influencing factors and i represents the corresponding influencing factor; a corresponding training set is established using the above method for training, and the output data is the impact calibration value PH(s). i This affects the calibration value to be 1 or 0;

[0063] By analyzing the impact data of the corresponding influencing factors through the impact calibration model, the corresponding impact calibration values ​​are obtained.

[0064] When the influence calibration value is 1, the influence judgment result is that there is an influence;

[0065] When the influence calibration value is 0, the influence judgment result is no influence.

[0066] In one embodiment, the pool area is divided into zones based on target factors. This can be done using existing methods such as clustering algorithms, deep learning algorithms, and machine learning to form different segmented regions. Each segmented region has the same impact on the identification of the VTN model.

[0067] In one embodiment, dividing the pool area into zones based on target factors includes:

[0068] Step SA1: Generate a pool area map of the pool area, obtain the influence identification features corresponding to the target factors, send the influence identification features of the target factors to the user's client, and receive the influence material data sent by the user's client; mark the feature identification data of the influence factors that have an impact on the corresponding locations in the pool area map according to the influence material data;

[0069] Step SA2: Merge adjacent locations with identical feature recognition data in the pool area map to obtain several unit areas;

[0070] Step SA3: Evaluate whether adjacent unit regions meet the merging requirements, merge adjacent unit regions that meet the merging requirements, and obtain new unit regions;

[0071] Step SA4: Repeat step SA3 until there are no adjacent cell regions that meet the merging requirements, then mark the cell region as a segmented region; mark the current pool area map as a regional distribution map.

[0072] In one embodiment, when there are multiple different feature recognition data for the same location in the pool area map, the feature recognition data that has the greatest impact on drowning identification in the VTN model is retained.

[0073] In one embodiment, assessing whether adjacent unit regions meet the merging requirements includes:

[0074] The feature recognition data corresponding to the identification unit area is used to determine its impact on the accuracy of drowning identification in the VTN model. Statistics are performed based on the corresponding historical recognition data, such as taking the average, mode, and other statistical values ​​of the many factors affecting accuracy. The corresponding statistical values ​​are marked as the impact values. If the unit area has multiple feature recognition data due to merging, the feature recognition data with the greatest impact is used.

[0075] Calculate the absolute value of the difference between the corresponding influence values ​​of adjacent unit regions and mark it as the influence difference;

[0076] When the difference in impact is greater than the threshold X1, the adjacent unit regions are not evaluated as not meeting the merging requirements.

[0077] When the difference in impact is not greater than the threshold X1, the adjacent unit regions are evaluated to meet the merging requirements.

[0078] The threshold X1 is set by those skilled in the art based on the actual situation or obtained through simulation of a large amount of data, such as by the platform provider based on whether differential analysis is needed during the establishment of the VTN model.

[0079] The model building module is used to build a VTN model based on the regional distribution map, adjust the regional distribution map according to the VTN model to obtain a baseline distribution map, and send the baseline distribution map to the pool information module of the corresponding user terminal; and deploy the VTN model in the risk identification module of the corresponding user terminal.

[0080] In one embodiment, a VTN model is built based on a regional distribution map. This model is built using existing VTN (VideoTransformer Networks) technology and is tailored to the actual situation of the regional distribution map. Figure 2 The diagram shows an example of the technical roadmap for the VTN model.

[0081] For example, data preprocessing and segmentation:

[0082] Acquire swimming pool videos, covering different lighting and water quality (emphasizing target factors), capturing normal swimming (breaststroke / freestyle / butterfly / backstroke) and drowning actions (pool drowning / child drowning). Collect 200 videos for each swimming style, with each video lasting approximately 5-8 seconds; for both pool drowning and child drowning behaviors, collect 100 videos for each drowning action, with each video lasting 5-8 seconds.

[0083] Dividing the video into fixed-length segments (e.g., 16 frames per segment) reduces the computational complexity of a single inference. Common segmentation methods include the sliding window method (dividing the video into non-overlapping or partially overlapping segments) and the uniform sampling method (uniformly sampling a fixed number of frames from a long video, such as sampling 5 frames per second to form multiple segments).

[0084] The segmented frame data undergoes spatial normalization, standardization, inference augmentation (using only center-based cropping to avoid random augmentation during training, such as flipping or multi-scale cropping), and temporal alignment. Then, the temporal and batch dimensions are merged to adapt to the input of the 2D backbone network. The data format is (batch_size×T,C,H,W). For example, if the original video is sampled as 16 frames, each frame is scaled to 224×224 resolution, forming an input tensor of shape (4,16,3,224,224).

[0085] Model structure construction:

[0086] 2D Spatial Feature Extraction: Spatial features are extracted frame by frame using a 2D backbone network (such as ResNet or ViT). Different backbone networks or network parameters can be selected or adjusted based on the differences in influencing factors for different areas of the pool. For example, for areas heavily affected by reflections, a network structure with strong robustness to changes in illumination can be selected, determined based on the regional distribution map.

[0087] Temporal attention encoding: Combining the sliding window attention mechanism of Longformer, the temporal dimension is modeled with linear complexity (O(n)), and keyframe information is captured through global attention marked with [CLS]. When processing pool videos, the weights or parameters of temporal attention encoding can be adjusted according to the degree of influence of different regions to highlight the features of key areas.

[0088] Classification MLP Header: Used to obtain the final class prediction.

[0089] Model optimization considering factors affecting swimming pools:

[0090] Adjust the model according to the influencing factors in different regions:

[0091] Reflective areas: If certain areas are severely affected by water surface reflection, a de-reflection algorithm can be added during the data preprocessing stage, or data augmentation of reflective samples can be added during model training to improve the model's adaptability to reflective scenes.

[0092] Occlusion areas: For areas where people frequently obstruct the view, a multi-view fusion method can be introduced into the model, combining video input from multiple cameras to reduce the impact of occlusion on behavior recognition.

[0093] Varying lighting conditions: In outdoor swimming pools, lighting conditions may vary in different areas. The exposure compensation parameters of the model can be adjusted or an adaptive lighting algorithm can be added based on the degree of lighting variation.

[0094] Data Acquisition and Labeling: When collecting pool video data, it is essential to fully consider the influencing factors in different areas to ensure that the dataset covers various scenarios. Simultaneously, when labeling the data, it is crucial to accurately label the areas where drowning behavior occurs and the influencing factors so that the model can learn the behavioral characteristics of different areas.

[0095] Model training and evaluation:

[0096] Training process: The VTN model is trained using a pre-collected and labeled dataset. During training, transfer learning can be employed, using model parameters pre-trained on large image or video datasets for initialization to accelerate model convergence.

[0097] Evaluation metrics: In addition to commonly used metrics such as accuracy and recall, metrics such as regional sensitivity can be introduced to evaluate the model's performance in different regions, taking into account the specific characteristics of swimming pool drowning risk identification. Based on the evaluation results, the model can be optimized and adjusted.

[0098] In one embodiment, the regional distribution map is adjusted according to the VTN model. Due to the differences in VTN models, it is necessary to determine the influence value of the current VTN model at each location. This is done by using relevant historical data for verification, or by statistically analyzing the verification results during the VTN model establishment process to determine the influence value at each location. The regions are then segmented and adjusted based on the influence values ​​at each location, and the influence values ​​corresponding to each segmented region are supplemented. The adjusted regional distribution map is then marked as the baseline distribution map.

[0099] For example, feature recognition data that may exist at each location in the regional distribution map is obtained, and corresponding verification data is generated based on the feature recognition data. The verification data includes simulated monitoring data that meets the feature recognition data and real standard result data. The verification data is analyzed through the VTN model to obtain the influence value of the VTN model at the corresponding location, and the largest influence value is selected as the influence value of that location. The segmented regions are adjusted according to the influence values ​​of each location, and the influence values ​​corresponding to each segmented region are supplemented.

[0100] The user terminal includes a pool information module, a monitoring module, a risk identification module, and an area display module.

[0101] The swimming pool information module is used to collect swimming pool information, send the swimming pool information to the swimming pool module on the platform, and collect influence material data based on the influence identification features of the received target factors. The influence material data consists of corresponding feature identification data and is sent to the swimming pool module on the platform. The received baseline distribution map is displayed to the user so that the user can have an intuitive understanding of the risk identification situation in the swimming pool area.

[0102] The impact data is collected based on the impact identification features. That is, the corresponding historical data is identified based on the impact identification features to obtain the corresponding feature identification data. For example, for the impact identification features of reflection, the historical monitoring data is identified based on the impact identification features to determine which locations have reflection and the intensity of reflection and other related feature data.

[0103] The monitoring module is used to monitor the swimming pool in real time, obtain the corresponding monitoring data, and send the monitoring data to the risk identification module.

[0104] The risk identification module is used to perform real-time analysis of monitoring data based on the VTN model to obtain corresponding risk identification results; and to perform corresponding processing based on the risk identification results, such as various early warning measures, specifically according to the user's management needs.

[0105] The area display module is used to analyze the risk identification status of the pool area, acquire monitoring data, identify feature recognition data that affects the drowning identification of the VTN model based on the monitoring data, and determine the influence value of the VTN model under the corresponding feature recognition data. Initially, it can be matched based on the influence value of the VTN model under the corresponding feature recognition data verified and statistically analyzed by the platform. Subsequently, it can be used to calculate the accuracy of the VTN model under the corresponding feature recognition data based on the recognition data accumulated by the user, and then determine the influence value based on the accuracy under normal conditions. The influence value is marked on the preset pool area map, and the pool area map is merged according to the influence value of each location to obtain several segmented areas. The influence value of the normal location area is 0. The segmented areas are merged according to the segmentation method in the above embodiment. The corresponding influence value is marked for the segmented areas. The pool area map is marked as a risk situation map and displayed to the pool management personnel.

[0106] By dynamically displaying risk situation maps to pool-related personnel, user managers can more effectively manage pool risks, strengthen the management of high-impact areas, improve drowning detection accuracy, and compensate for potential intelligent recognition errors.

[0107] The above formulas are all numerical calculations after removing dimensions. The formulas are obtained by software simulation based on a large amount of data and are closest to the real situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulation based on a large amount of data.

[0108] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A swimming pool environment drowning risk identification and control system based on the VTN model, characterized in that, Including both the platform side and the user side; The platform includes a pool module and a model building module; the user terminal includes a pool information module, a monitoring module, and a risk identification module. The pool module is used to perform pool analysis, obtain pool information sent by the corresponding user terminal, and obtain impact analysis data of the preset influencing factors based on the pool information; Based on the impact analysis data, determine whether the swimming pool is affected by the influencing factors, and obtain the impact judgment result of the influencing factors, which includes whether there is an impact or no impact; mark the influencing factors with the impact judgment result of having an impact as target factors; The pool area is divided into zones based on the corresponding target factors to obtain a regional distribution map of the target factors. The model building module is used to build a VTN model based on the regional distribution map, adjust the regional distribution map according to the VTN model to obtain a corresponding baseline distribution map, send the regional distribution map to the pool information module of the corresponding user terminal, and deploy the VTN model into the risk identification module of the corresponding user terminal. The pool information module is used to collect pool information, send the pool information to the pool module on the platform, and collect influence material data based on the influence identification features of the received target factors. The influence material data consists of feature identification data and is sent to the pool module on the platform. The received baseline distribution map will be displayed to the user; The monitoring module is used to monitor the swimming pool in real time, obtain the corresponding monitoring data, and send the monitoring data to the risk identification module. The risk identification module is used to perform real-time analysis of monitoring data based on the VTN model, obtain corresponding risk identification results, and perform corresponding processing based on the risk identification results. The regional distribution map was adjusted based on the VTN model, including: Obtain feature recognition data for each location in the regional distribution map, and generate corresponding verification data based on the feature recognition data; The VTN model was used to analyze the validation data to obtain the influence value of the VTN model at the corresponding locations. The segmented regions in the regional distribution map are adjusted according to the influence values ​​of each location, and the influence values ​​corresponding to each segmented region are added. The adjusted regional distribution map is then marked as the baseline distribution map. The user terminal also includes a region display module, which is used to perform risk identification and status analysis on the pool area, obtain a risk situation map, and display the risk situation map to the pool management personnel. Risk identification and status analysis of the swimming pool area, including: Acquire monitoring data, identify the feature data that affects drowning identification in the VTN model based on the monitoring data, and determine the influence value of the VTN model under the corresponding feature data. Based on the influence values ​​of each location, the pool area map is merged to obtain several segmented areas, and the corresponding influence values ​​are marked for each segmented area; the pool area map is then marked as a risk situation map.

2. The swimming pool environment drowning risk identification and control system based on the VTN model according to claim 1, characterized in that, Based on the impact analysis data, determine whether the swimming pool is affected by the corresponding influencing factors, including: The platform provides an impact identification library, which stores the impact identification features corresponding to each influencing factor. An impact calibration model is established based on the impact identification library. The expression for the impact calibration model is as follows: ； In the formula: s i This represents the impact analysis data of the corresponding influencing factors, where i represents the corresponding influencing factor, i = 1, 2, ..., n, and n is the number of influencing factors; the output data is the impact calibration value PH(s). i This affects the calibration value to be 1 or 0; The influence calibration value of the influencing factors is obtained by analyzing the influence analysis data of the corresponding influencing factors through the influence calibration model. When the influence calibration value is 1, the influence judgment result of the influencing factor is that it has an influence; When the influence calibration value is 0, the influence of the influencing factor is judged as having no influence.

3. The swimming pool environment drowning risk identification and control system based on the VTN model according to claim 1, characterized in that, The pool area is divided into zones based on target factors, including: Step SA1: Generate a pool area map of the pool area, obtain the influence identification features of the target factors, send the influence identification features of the target factors to the user terminal of the corresponding user, and receive the influence material data sent by the corresponding user terminal; mark the feature identification data of the influence factors that have an impact on the corresponding location in the pool area map according to the influence material data; Step SA2: Merge adjacent locations with identical feature recognition data in the pool area map to obtain several unit areas; Step SA3: Evaluate whether adjacent unit regions meet the merging requirements, merge adjacent unit regions that meet the merging requirements, and obtain new unit regions; Step SA4: Repeat step SA3 until there are no adjacent cell regions that meet the merging requirements, then mark the cell region as a segmented region; mark the current pool area map as a regional distribution map.

4. The swimming pool environment drowning risk identification and control system based on the VTN model according to claim 3, characterized in that, Step SA3 assesses whether adjacent cell regions meet the merging requirements, including: Identify the feature recognition data corresponding to the unit region, and determine the influence value of the unit region based on the feature recognition data; Calculate the absolute value of the difference between the corresponding influence values ​​of adjacent unit regions and mark it as the influence difference; The impact difference is used to determine whether the merging requirements between adjacent unit areas are met.

5. The swimming pool environment drowning risk identification and control system based on the VTN model according to claim 4, characterized in that, When there are multiple different feature recognition data for the same location or the same unit area in the pool area map, retain the feature recognition data that has the greatest impact on drowning identification in the VTN model.

Images