High-altitude operation anti-falling safety monitoring method based on data processing

Abstract

The invention relates to the technical field of high-altitude operation monitoring, and discloses a high-altitude operation anti-falling safety monitoring method based on data processing. The method comprises the steps of obtaining an environmental parameter set of a target operation area, and performing matching query in a historical accident database to determine a historical falling accident feature set; performing risk grade division on the falling accident feature set to obtain a plurality of graded historical falling accident feature sets, and traversing and extracting an environmental risk factor reference value and a risk influence range parameter; taking the risk influence range parameter as a boundary condition of a dynamic monitoring window, and performing multi-dimensional feature capture on the real-time motion data sequence of the high-altitude operation personnel by utilizing a plurality of configured dynamic monitoring windows to obtain a plurality of real-time motion feature sets; and carrying out space-time correlation fusion on the plurality of real-time motion feature sets to obtain a target space-time fusion motion feature set, and taking the target space-time fusion motion feature set as an anti-falling safety monitoring result. According to the method, environment and motion data are fused, and a precise monitoring basis is provided for high-altitude operation falling prevention.

Description

Technical Field

[0001] This invention relates to the field of high-altitude operation monitoring technology, specifically a high-altitude operation fall prevention safety monitoring method based on data processing. Background Technology

[0002] Working at heights is a crucial form of work in industries such as construction, power, and communications, and its safety has always been a major concern. Falls are the primary threat to the lives of workers. According to statistics, falls account for over 60% of all accidents involving working at heights, and the consequences are often severe. Despite increased safety oversight and upgrades to protective equipment in recent years, falls still occur frequently, which is closely related to the limitations of current safety monitoring technologies.

[0003] Among existing methods for monitoring safety in high-altitude operations, manual monitoring remains a common approach. However, manual monitoring is limited by the duration of a person's attention and judgment, making it prone to oversights in complex working environments or when the work period is long. Especially in situations where the high-altitude work platform is swaying or visibility is low, it is difficult for humans to accurately capture subtle changes in the workers' movements, often only reacting after an accident has occurred, thus failing to provide early warnings.

[0004] With the application of sensor technology, some monitoring systems have begun to use devices such as accelerometers and gyroscopes to collect motion data of workers. However, these systems often focus on monitoring single parameters, such as judging whether a fall has occurred solely based on changes in acceleration, while neglecting the impact of environmental factors on work safety. In reality, the risk level of working at heights is closely related to environmental parameters. For example, excessive wind speed may cause workers to lose their balance, high humidity may cause the work platform to slip, and abnormal temperature may affect the workers' physical condition. Existing systems lack integrated analysis of these environmental parameters, resulting in incomplete risk assessments.

[0005] Some intelligent monitoring systems have attempted to introduce data processing algorithms, but significant shortcomings remain in data association and fusion. These systems fail to effectively utilize historical accident data and cannot compare and analyze the characteristics of the current working environment with those of historical fall accidents, resulting in a lack of effective reference for risk assessment. For example, under the same working height and equipment conditions, different weather conditions may correspond to different fall risk probabilities, but existing systems cannot quantify such risks based on historical data. Monitoring of worker movement data is also limited, typically focusing only on basic parameters such as position and speed, while ignoring detailed features such as continuous changes in limb posture and movement trajectory, making it difficult to accurately identify potential danger signals such as subtle swaying before loss of balance or slipping.

[0006] Existing monitoring systems primarily process data within a single spatiotemporal dimension, failing to achieve temporal and spatial correlation and fusion of motion and environmental data. The formation of risks in high-altitude operations is often a dynamic process; environmental changes at a particular moment can trigger abnormal motion states over a subsequent period. Existing systems struggle to capture this spatiotemporal correlation, resulting in delayed or incomplete monitoring results that fail to reflect the evolving risk process in a timely manner. These technical deficiencies make it difficult for existing monitoring methods to provide comprehensive and accurate safety monitoring results when dealing with complex and ever-changing high-altitude working environments, and thus, they are unable to effectively prevent fall accidents. Summary of the Invention

[0007] The purpose of this invention is to provide a high-altitude operation fall prevention safety monitoring method based on data processing, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides a high-altitude operation fall prevention safety monitoring method based on data processing, the method comprising:

[0009] Obtain the set of environmental parameters for the target work area, and perform a matching query in the historical accident database based on the set of environmental parameters to determine the set of historical fall accident characteristics.

[0010] The risk level of the historical fall accident feature set is classified to determine multiple graded historical fall accident feature sets;

[0011] By traversing the multiple sets of historical fall accident features of different levels, environmental risk factors are extracted to obtain multiple environmental risk factor benchmark values ​​and multiple risk impact range parameters.

[0012] The multiple risk impact range parameters are used as boundary conditions for multiple dynamic monitoring windows. The configured multiple dynamic monitoring windows are used to capture multi-dimensional features of the real-time motion data sequence of high-altitude workers, and multiple sets of real-time motion features are obtained.

[0013] The multiple real-time motion feature sets are spatiotemporally correlated and fused to obtain a target spatiotemporally fused motion feature set, which is then used as the fall prevention safety monitoring result.

[0014] Preferably, environmental risk factors are extracted by traversing the multiple sets of historical fall accident characteristics at different levels, resulting in multiple environmental risk factor baseline values ​​and multiple risk impact range parameters, including:

[0015] By traversing the multiple sets of historical fall accident features of different levels, environmental parameters are extracted to obtain multiple sets of historical environmental parameter sequences of different levels.

[0016] Critical environmental parameters are extracted from the multiple sets of graded historical environmental parameter sequences to determine multiple sets of historical critical environmental parameters. Each historical critical environmental parameter is an environmental parameter in each graded historical environmental parameter sequence that is close to the threshold for the occurrence of a fall accident.

[0017] Factor extraction is performed on the multiple sets of historical critical environmental parameters to obtain multiple sets of historical environmental risk factors, and statistical analysis is performed on the multiple sets of historical environmental risk factors to determine the benchmark values ​​of multiple environmental risk factors.

[0018] Using the multiple sets of historical environmental risk factors as a reference, the risk impact range of the multiple sets of hierarchical historical environmental parameter sequences is defined, and multiple sets of historical risk impact ranges are determined. Each historical risk impact range reflects the duration range of environmental parameters associated with critical environmental parameters in a hierarchical historical environmental parameter sequence.

[0019] Calculate the span value of each of the multiple historical risk impact range sets to obtain the multiple risk impact range parameters.

[0020] Preferably, factor extraction is performed on the multiple sets of historical critical environmental parameters to obtain multiple sets of historical environmental risk factors, and statistical analysis is performed on the multiple sets of historical environmental risk factors to determine multiple environmental risk factor benchmark values, including:

[0021] The risk factor extraction algorithm is used to extract factors from the multiple sets of historical critical environmental parameters to obtain the multiple sets of historical environmental risk factors.

[0022] The median of multiple historical environmental risk factors is calculated by traversing multiple sets of historical environmental risk factors;

[0023] According to a preset step size, the median of the multiple historical environmental risk factors is iteratively optimized in the set of multiple historical environmental risk factors to obtain multiple iterative historical environmental risk factors.

[0024] When the coefficient of variation of the plurality of iterative historical environmental risk factors is less than or equal to the coefficient of variation of the median of the plurality of historical environmental risk factors, the median of the plurality of historical environmental risk factors is used as the benchmark value of the plurality of environmental risk factors.

[0025] Preferably, the method described includes:

[0026] When the coefficient of variation of the plurality of iterated historical environmental risk factors is greater than the coefficient of variation of the median of the plurality of historical environmental risk factors, it is determined whether the difference between the coefficient of variation of the plurality of iterated historical environmental risk factors and the median of the plurality of historical environmental risk factors is greater than or equal to a preset coefficient of variation difference threshold. If so, the iteration is continued based on the plurality of iterated historical environmental risk factors until the maximum number of iterations is met, and the plurality of iterated historical environmental risk factors obtained in the last iteration are used as the benchmark values ​​of the plurality of environmental risk factors.

[0027] If not, then stop the iteration and use the multiple historical environmental risk factors as the benchmark values ​​for the multiple environmental risk factors.

[0028] Preferably, using the multiple sets of historical environmental risk factors as a reference, the risk impact range of the multiple sets of hierarchical historical environmental parameter sequences is defined to determine multiple sets of historical risk impact ranges, including:

[0029] The risk factor extraction algorithm is used to extract environmental features from the multiple sets of hierarchical historical environmental parameter sequences to obtain multiple sets of hierarchical historical environmental feature sequences.

[0030] One historical environmental risk factor is randomly extracted from the multiple sets of historical environmental risk factors as the first historical environmental risk factor, and the corresponding first-level historical environmental feature sequence is matched from the multiple sets of hierarchical historical environmental feature sequences.

[0031] According to a preset association threshold, the first historical environmental risk factor is retrieved in the first hierarchical historical environmental feature sequence to obtain the first historical environmental risk factor interval.

[0032] The time span of the first historical environmental risk factor interval is statistically analyzed, and the statistical results are used as the scope of the first historical risk impact.

[0033] According to a preset correlation threshold, the risk impact range of the multiple sets of historical environmental risk factors is defined in the corresponding sets of multiple hierarchical historical environmental parameter sequences, thereby determining multiple sets of historical risk impact ranges.

[0034] Preferably, the multiple real-time motion feature sets are spatiotemporally correlated and fused to obtain a target spatiotemporally fused motion feature set, including:

[0035] Randomly extract a first real-time motion feature set and a second real-time motion feature set from the plurality of real-time motion feature sets;

[0036] Calculate the spatiotemporal correlation between the first real-time motion feature set and the second real-time motion feature set to determine the first spatiotemporal correlation set;

[0037] The first spatiotemporal correlation set is standardized to obtain the first spatiotemporal correlation standardized value set.

[0038] Perform matrix operations on the first set of standardized spatiotemporal correlation values ​​and the second set of real-time motion features to obtain the first set of spatiotemporal fused motion features.

[0039] A third real-time motion feature set is randomly extracted from the plurality of real-time motion feature sets, and spatiotemporally correlated and fused with the first spatiotemporal fusion motion feature set to obtain a second spatiotemporal fusion motion feature set;

[0040] After multiple spatiotemporal correlation fusions, until all real-time motion features in the multiple real-time motion feature sets are fused, the target spatiotemporal fused motion feature set is obtained.

[0041] Preferably, the method described includes:

[0042] Multiple sets of standardized values ​​of spatiotemporal correlation of samples, multiple sets of real-time motion features of samples, and multiple sets of spatiotemporal fusion motion features of samples are obtained as training samples;

[0043] The network model based on the recurrent neural network is trained under supervision using training samples until it converges, thus obtaining the trained spatiotemporal fusion network model.

[0044] The first spatiotemporal fusion motion feature set is obtained by performing matrix operations on the first spatiotemporal correlation standardized value set and the second real-time motion feature set using the spatiotemporal fusion network model.

[0045] Preferably, the historical fall accident feature set is classified into risk levels to determine multiple graded historical fall accident feature sets, including:

[0046] Multiple historical fall accident features are randomly selected from the set of historical fall accident features;

[0047] By comparing the characteristics of the multiple historical fall accidents in pairs, multiple comparison combinations are obtained;

[0048] Determine whether there are any comparison combinations whose feature similarity is less than a preset feature similarity threshold. If not, then use the features of the multiple historical fall accidents as multiple grading benchmarks.

[0049] Based on the multiple grading benchmarks, the risk levels of the historical fall accident feature set are divided according to a preset feature similarity threshold to obtain the multiple graded historical fall accident feature sets, wherein each graded historical fall accident feature set corresponds to a grading benchmark.

[0050] Preferably, after using the target spatiotemporal fusion motion feature set as the fall safety monitoring result, the method further includes:

[0051] The deviation of the target motion features is determined by comparing the target spatiotemporal fusion motion feature set with the preset safety threshold set.

[0052] When the deviation of the target motion feature is greater than a preset deviation threshold, a fall prevention warning signal is generated. The fall prevention warning signal includes a warning level and a warning trigger location.

[0053] The fall prevention warning signal is transmitted to the high-altitude operation monitoring terminal, which then performs an audible and visual alarm operation based on the fall prevention warning signal.

[0054] Preferably, transmitting the fall prevention warning signal to the high-altitude work monitoring terminal includes:

[0055] The fall protection warning signal is encrypted to obtain an encrypted warning signal. The encryption process uses a symmetric encryption algorithm.

[0056] Establish a wireless communication connection with the high-altitude operation monitoring terminal;

[0057] The encrypted warning signal is sent to the high-altitude operation monitoring terminal via the wireless communication connection. The high-altitude operation monitoring terminal decrypts the received encrypted warning signal and then performs an audible and visual alarm operation.

[0058] Compared with the prior art, the beneficial effects of the present invention are:

[0059] The high-altitude operation fall prevention safety monitoring method based on data processing provided by this invention addresses the shortcomings of existing technologies by forming a more comprehensive and accurate monitoring logic, and its effects are reflected in multiple aspects.

[0060] This method acquires a set of environmental parameters for the target work area and matches them with a historical accident database. It establishes a direct link between the current work environment and historical fall accident characteristics, making risk identification no longer limited to single real-time data but incorporating accumulated historical experience. The historical fall accident characteristic set contains key elements of fall accidents under different environmental conditions, such as accident triggers under specific wind speed and humidity combinations, and typical movement characteristics of workers before a hazard occurs. Through this matching, the system can quickly locate potential risk types from current environmental parameters. For example, when a query reveals multiple historical slippage and fall accidents under a certain humidity and platform material combination, the system can focus on monitoring the movement characteristics of workers' feet in advance, avoiding the blind risk judgment caused by the lack of historical data support in existing methods.

[0061] Classifying the risk level of historical fall accident characteristics allows for the refinement of complex risk factors according to severity and scope of impact, making the subsequently extracted environmental risk factor benchmark values ​​more targeted. Different risk levels correspond to different scopes of impact and characteristic manifestations. After classification, the extracted risk impact range parameters can more accurately define the areas and dimensions requiring key monitoring. For example, high-risk levels may correspond to a larger monitoring range and more motion characteristic dimensions, while low-risk levels can have a smaller monitoring range to reduce data redundancy. This differentiated processing solves the problem of existing systems having fixed monitoring ranges and being unable to adapt to different risk levels, allowing for a more rational allocation of monitoring resources.

[0062] Multiple dynamic monitoring windows, using the risk impact range parameter as boundary conditions, capture multi-dimensional features of real-time motion data sequences of high-altitude workers, overcoming the limitations of existing methods that rely on a single motion parameter. These dynamic monitoring windows can selectively capture features at multiple levels, such as limb swing angles, the continuity of movement trajectories, and the frequency of speed changes, depending on the risk impact range. For example, when the risk impact range involves the edge area of ​​the work platform, the monitoring window can focus on capturing changes in the worker's position and body tilt angle; when the risk impact range is related to wind speed, the monitoring window can simultaneously monitor the worker's balance adjustments and the stability of their movement speed. This multi-dimensional capture comprehensively reflects the real-time status of the workers, avoiding the omission of critical risk signals due to monitoring a single parameter.

[0063] By performing spatiotemporal correlation fusion on multiple real-time motion feature sets, scattered motion data is integrated across time and space dimensions. The resulting target spatiotemporal fused motion feature set can more realistically reflect the dynamic evolution of risks. Risks in high-altitude operations are often not isolated; environmental changes at a certain moment may trigger a series of abnormal motion states. For example, a sudden gust of wind may cause workers to tilt more sharply after 10 seconds. Existing methods struggle to capture such spatiotemporal correlations. This method, however, by fusing motion features and environmental parameters from different times and locations, can identify complex risk characteristics such as "under a specific wind speed, a worker moves towards the edge of the platform for 5 consecutive seconds with gradually increasing upper limb swing amplitude." This dynamic correlation analysis makes the monitoring results more closely match the actual process of risk formation, reducing misjudgments or omissions caused by fragmented information. Attached Figure Description

[0064] Figure 1 This is a schematic diagram illustrating the working principle of the high-altitude operation fall prevention safety monitoring method based on data processing described in this invention.

[0065] Figure 2A flowchart for extracting environmental risk factors and determining parameters for the scope of risk impact;

[0066] Figure 3 A flowchart for defining the scope of risk impact;

[0067] Figure 4 A flowchart for transmitting fall warning signals. Detailed Implementation

[0068] 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.

[0069] Please see Figure 1 This invention provides a high-altitude operation fall prevention safety monitoring method based on data processing, the method comprising:

[0070] A set of environmental parameters for the target work area is obtained. These parameters, including wind speed, light intensity, temperature, humidity, and ground flatness, reflect the working environment's condition and can be collected in real-time by a sensor array deployed in the work area. Based on this set of environmental parameters, a matching query is performed in a historical accident database. This database stores information such as environmental parameters and accident characteristics corresponding to past falls from heights. A similarity matching algorithm is used to find historical records similar to the current set of environmental parameters, thereby determining a set of historical fall accident characteristics. This set includes features such as environmental change trends and personnel movement at the time of the accident.

[0071] The risk level of the historical fall accident feature set is classified based on factors such as the severity of the accident consequences and the probability of occurrence. The feature set is divided into multiple graded historical fall accident feature sets by clustering algorithms or preset thresholds, with each set corresponding to a different risk level.

[0072] Environmental risk factors were extracted by traversing multiple sets of historical fall accident characteristics across different levels, resulting in multiple environmental risk factor benchmark values ​​and multiple risk impact range parameters. The environmental risk factor benchmark values ​​serve as reference values ​​reflecting the critical state of environmental risk, while the risk impact range parameters reflect the temporal or spatial span of the risk's impact.

[0073] Multiple risk impact range parameters are used as boundary conditions for multiple dynamic monitoring windows. The size and monitoring dimensions of the dynamic monitoring windows are adaptively adjusted according to the parameters. For example, the window length in the time dimension is determined by the time span of the risk impact, and the monitoring range in the spatial dimension is determined by the spatial range of the risk impact. Multiple configured dynamic monitoring windows are used to capture multi-dimensional features of real-time motion data sequences of high-altitude workers. The real-time motion data sequences include the worker's position coordinates, velocity, acceleration, and posture angle, collected through wearable devices or visual sensors. The captured sets of multiple real-time motion features cover different dimensions of motion state information.

[0074] Multiple real-time motion feature sets are spatiotemporally correlated and fused. The fusion process needs to consider the temporal sequence and spatial positional relationship of different features, eliminate data redundancy and strengthen key features, and finally obtain the target spatiotemporally fused motion feature set. This set comprehensively reflects the overall motion state of personnel in the spatiotemporal dimension. It can be used as the result of fall prevention safety monitoring and can intuitively present the fall risk status faced by the current workers.

[0075] Example 1: See Figure 2 The process of extracting environmental risk factors by traversing multiple sets of historical fall accident characteristics at different levels, and obtaining multiple environmental risk factor benchmark values ​​and multiple risk impact range parameters, requires multiple consecutive processing steps.

[0076] Environmental parameters are extracted by traversing multiple graded historical fall incident feature sets. Each graded historical fall incident feature set corresponds to a specific risk level and contains detailed information related to all historical fall incidents under that level. When extracting environmental parameters, various types of environmental data related to the incident need to be extracted from each graded set. This data covers the environmental state before, during, and after the incident, ultimately forming multiple graded historical environmental parameter sequence sets. Each graded historical environmental parameter sequence set consists of several graded historical environmental parameter sequences, each sequence recording the changes in environmental parameters corresponding to a specific historical incident in chronological order. For example, a sequence might sequentially record wind speed changes from 10 minutes before the incident to the time of the incident.

[0077] Critical environmental parameters are extracted from multiple sets of graded historical environmental parameter sequences. Critical environmental parameters refer to those environmental parameters in each graded historical environmental parameter sequence that are close to the fall accident threshold. The corresponding fall accident thresholds differ for different types of environmental parameters. For example, the wind speed threshold might be set to 10 m / s; when the wind speed value in the sequence reaches 9.5 m / s, this value is considered a critical environmental parameter. During the extraction process, each graded historical environmental parameter sequence is analyzed one by one. By comparing each parameter value in the sequence with the preset accident threshold, parameters that meet the criteria are selected, thereby determining multiple sets of historical critical environmental parameters. Each set contains all critical environmental parameters from the corresponding graded historical environmental parameter sequence set.

[0078] Factor extraction is performed on multiple historical critical environmental parameter sets. Using a risk factor extraction algorithm, representative environmental risk factors are extracted from each historical critical environmental parameter set. These factors reflect the key environmental factors leading to fall accidents. For example, from a set containing critical parameters such as wind speed and humidity, strong wind factors and high humidity factors may be extracted, thus obtaining multiple sets of historical environmental risk factors. Subsequently, statistical analysis is performed on these sets of historical environmental risk factors, analyzing the frequency and numerical distribution of each risk factor across different historical records. By analyzing these statistical results, benchmark values ​​for multiple environmental risk factors are determined. These benchmark values ​​summarize the typical values ​​of various environmental risk factors under critical conditions. For example, the benchmark value for the strong wind factor might be a reference value determined by statistically analyzing the concentrated distribution range of critical wind speeds in multiple fall accidents caused by strong winds.

[0079] Using multiple sets of historical environmental risk factors as references, the risk impact range of multiple graded historical environmental parameter sequence sets is defined. For each sequence in each graded historical environmental parameter sequence set, environmental parameters associated with critical environmental parameters are identified based on the factor characteristics in the historical environmental risk factor set. These associated environmental parameters constitute a continuous interval, which is the historical risk impact range. It reflects the range of environmental parameter changes in the graded historical environmental parameter sequence from the initial manifestation of the risk factor's impact to its disappearance. For example, in a sequence associated with strong wind factors, the risk impact range might be the interval consisting of all wind speed parameters from the time the wind speed first exceeds 5 m / s to the time it finally drops below 5 m / s, thereby determining multiple sets of historical risk impact ranges.

[0080] Calculate the span values ​​for multiple sets of historical risk impact ranges. The calculation of the span value needs to be determined according to the type of risk impact range. If it is a continuous time interval, the span value is the time length of that interval, such as 30 minutes; if it is a spatial distribution range, the span value is the spatial distance of that range, such as 5 meters. By calculating the span values ​​of all intervals in each set of historical risk impact ranges, multiple risk impact range parameters are finally obtained. These parameters quantify the size of the impact range of environmental risk factors under different risk levels.

[0081] Example 2: See Figure 3 In determining the baseline values ​​of environmental risk factors based on multiple sets of historical critical environmental parameters, when the coefficient of variation of the iterative historical environmental risk factors is greater than the coefficient of variation of the median of the historical environmental risk factors, it is necessary to further analyze the relationship between the difference in the coefficients of variation between the two and a preset threshold for the difference in the coefficient of variation. The coefficient of variation is calculated by the ratio of the standard deviation to the mean and is used to reflect the degree of dispersion of the data.

[0082] If the difference in the coefficients of dispersion between the two factors is greater than or equal to a preset threshold, iteration must continue based on the current set of multiple iterative historical environmental risk factors. During the iteration process, the relevant parameters for factor extraction will be readjusted, and factor extraction will be performed again on the set of historical critical environmental parameters to generate new iterative historical environmental risk factors. This process will be repeated until the preset maximum number of iterations is reached. The maximum number of iterations can be set according to the needs of the actual application scenario, for example, it can be set to 30 times. When the maximum number of iterations is reached, the multiple iterative historical environmental risk factors obtained in the last iteration are determined as the benchmark values ​​for multiple environmental risk factors.

[0083] If the difference in the coefficients of dispersion between the two is less than the preset threshold for the difference in coefficients of dispersion, the iteration operation is stopped, and the current multiple historical environmental risk factors are directly used as the benchmark values ​​for multiple environmental risk factors.

[0084] When defining the risk impact range of multiple hierarchical historical environmental parameter sequence sets by using multiple sets of historical environmental risk factors as references, the first step is to use a risk factor extraction algorithm to extract environmental features from the multiple hierarchical historical environmental parameter sequence sets. The extracted environmental features cover the key information in the parameter sequences that reflects environmental change trends, abrupt change points, etc., thus forming multiple hierarchical historical environmental feature sequence sets.

[0085] One historical environmental risk factor is randomly selected from multiple sets of historical environmental risk factors as the first historical environmental risk factor, for example, the factor of "instantaneous wind speed exceeding 15 m / s". Then, the first-level historical environmental feature sequence corresponding to the first historical environmental risk factor is matched from multiple sets of hierarchical historical environmental feature sequences. This sequence contains environmental feature information related to the selected risk factor.

[0086] According to a preset association threshold, a range search is performed on the first historical environmental risk factor in the first-level historical environmental feature sequence. The preset association threshold is a standard used to determine whether environmental features and risk factors are associated. When the degree of association between a certain segment of environmental features and the first historical environmental risk factor in the sequence reaches or exceeds the threshold, that segment is determined to be the range of the first historical environmental risk factor.

[0087] The time span of the first historical environmental risk factor interval is calculated from the start time to the end time of the interval, and the result is the scope of the first historical risk impact.

[0088] Following the same process described above, and using a preset correlation threshold as a standard, the risk impact range of each factor in multiple sets of historical environmental risk factors is sequentially defined within the corresponding sets of multiple hierarchical historical environmental parameter sequences. That is, for each risk factor, the corresponding hierarchical historical environmental feature sequence is first matched, then interval retrieval and time span statistics are performed to ultimately determine multiple sets of historical risk impact ranges. Each range within each set of historical risk impact ranges corresponds to the impact interval of different historical environmental risk factors within their respective environmental parameter sequences.

[0089] Example 3: Spatiotemporal correlation fusion of multiple real-time motion feature sets requires multiple rounds of feature integration. A first real-time motion feature set and a second real-time motion feature set are randomly selected from multiple real-time motion feature sets. The first real-time motion feature set may contain data such as the three-dimensional coordinates, movement speed, and angles of various joints of the worker at a certain moment, while the second real-time motion feature set may cover information such as acceleration changes, body tilt angles, and displacement trajectory segments at another moment.

[0090] The spatiotemporal correlation between the first real-time motion feature set and the second real-time motion feature set is calculated. The temporal correlation is obtained by analyzing the interval between the two sets on the time axis and whether the trend of change of motion parameters is consistent. For example, it is determined whether the velocity change at the previous moment is continuously related to the acceleration change at the next moment. The spatial correlation is determined by comparing the spatial distance of position coordinates and the degree of deviation of motion direction. Thus, the first spatiotemporal correlation set is formed. The elements in this set reflect the correlation strength between motion features of different dimensions.

[0091] The first spatiotemporal correlation set is standardized by converting all correlation values ​​to the range of 0 to 1. This standardized set eliminates the dimensional differences between different types of motion parameters, allowing for direct comparison and computation of correlation data from different dimensions. A matrix operation is then performed on the first spatiotemporal correlation standardized set and the second real-time motion feature set, using the following formula:

[0092] F1 = S1 × M2

[0093] Where F1 represents the first spatiotemporal fusion motion feature set, S1 represents the matrix corresponding to the first spatiotemporal correlation standardized value set, M2 represents the matrix corresponding to the second real-time motion feature set, and × represents matrix multiplication. This operation integrates the information from the two sets to generate the first spatiotemporal fusion motion feature set, which contains both the core information of the second real-time motion features and the correlation with the first real-time motion feature set.

[0094] A third set of real-time motion features is randomly extracted from multiple sets of real-time motion features. This set may contain data such as the worker's posture, gait frequency, and rate of displacement change in another time period. Using the same method as above, the spatiotemporal correlation between the third set of real-time motion features and the first spatiotemporal fusion motion feature set is calculated. By analyzing the connection between the two in the time series and the degree of correlation in spatial parameters, a new spatiotemporal correlation set is formed. After standardization, this new set of spatiotemporal correlation features is subjected to matrix operations with the first spatiotemporal fusion motion feature set to obtain the second spatiotemporal fusion motion feature set.

[0095] The above fusion process is repeated, with each iteration randomly selecting a set from the remaining real-time motion feature sets and performing spatiotemporal correlation analysis and matrix operations with the currently generated spatiotemporal fused motion feature set, gradually integrating new motion feature information. As the number of fusion iterations increases, the time span covered by the spatiotemporal fused motion feature set continuously expands, and the spatial dimension of motion features becomes more comprehensive. Ultimately, all features from multiple real-time motion feature sets are included in the fusion scope to form the target spatiotemporal fused motion feature set.

[0096] During matrix operations, a spatiotemporal fusion network model based on a recurrent neural network can be used. Training this model requires a large amount of sample data, including multiple sets of standardized spatiotemporal correlation values, multiple sets of real-time motion features, and multiple sets of spatiotemporal fusion motion features. These samples originate from historical monitoring data under different high-altitude work scenarios, covering motion features under various states such as normal operation, risk warning, and accident occurrence. The sample data is input into the network model for supervised training. By adjusting the network's weight parameters and activation function, the fusion result output by the model gradually approaches the sample spatiotemporal fusion motion feature sets. Training is complete when the model's output error stabilizes within a preset range. Using the trained spatiotemporal fusion network model to perform matrix operations on the first set of standardized spatiotemporal correlation values ​​and the second set of real-time motion features allows for more efficient processing of non-linearly correlated motion feature data, generating a first spatiotemporal fusion motion feature set that conforms to the actual scenario. Through this multi-round, iterative fusion approach, the final target spatiotemporal fusion motion feature set comprehensively reflects the motion state of high-altitude workers in different time and space dimensions, providing comprehensive feature data for fall prevention safety monitoring.

[0097] Example 4: The process of classifying the risk level of historical fall accident feature sets and determining multiple graded historical fall accident feature sets requires multiple feature extractions and comparisons. The historical fall accident feature sets contain a large number of detailed features of past high-altitude work fall accidents. These features cover various aspects such as environmental conditions, personnel operation behavior, and equipment status at the time of the accident. For example, the features of a certain accident might include "wind speed 8 m / s, working height 15 m, not wearing a safety belt, and loose scaffolding."

[0098] Multiple historical fall accident features are randomly selected from a historical fall accident feature set. The number of features selected depends on the total size of the set, typically choosing samples that cover different types of accidents, such as selecting features from 10 fall accidents in different scenarios. These selected historical fall accident features are then compared pairwise to form multiple comparison combinations, each containing two different historical fall accident features. During the comparison process, the similarity between the two features in dimensions such as environmental parameters, personnel behavior, and equipment status needs to be analyzed. For example, comparing whether the wind speed, working height, and personnel protective measures of the two accidents are similar, thereby obtaining the feature similarity of each comparison combination.

[0099] The system determines whether any of the feature similarity scores of multiple comparison combinations are below a preset feature similarity threshold. The preset feature similarity threshold is a standard used to distinguish whether features are similar. For example, if the threshold is set to 0.7, two features are considered similar if their similarity score is 0.8, and dissimilar if it is 0.6. If the feature similarity score of all comparison combinations is greater than or equal to the preset threshold, it indicates that the extracted features of multiple historical fall accidents have a high degree of consistency and can be used as multiple grading benchmarks.

[0100] Based on these grading benchmarks, the risk levels of historical fall accident feature sets are classified according to preset feature similarity thresholds. During the classification, the similarity between each feature in the historical fall accident feature set and each grading benchmark is calculated. Features with a similarity greater than or equal to a preset threshold to a given grading benchmark are assigned to the set corresponding to that benchmark, thus forming multiple graded historical fall accident feature sets. Each graded historical fall accident feature set corresponds to a grading benchmark, and all features within the set have a high similarity to that benchmark.

[0101] For example, if one of the classification criteria is "wind speed 7-9m / s, working height 10-20m, lack of protective measures", then during the classification process, all accident characteristics that are similar to this criterion to or exceed the preset threshold, such as "wind speed 8.5m / s, working height 18m, no safety belt worn" and "wind speed 7.2m / s, working height 15m, safety rope not secured", will be included in the classification set of historical fall accident characteristics corresponding to this criterion.

[0102] If there are comparison combinations with feature similarity less than a preset threshold, multiple historical fall accident features need to be re-extracted from the historical fall accident feature set, and the pairwise comparison process described above needs to be repeated. During re-extraction, the number or range of features extracted can be adjusted appropriately to avoid situations where feature differences are too large again. After multiple extractions and comparisons, until it is determined that there are no comparison combinations with feature similarity less than the preset threshold, the multiple historical fall accident features at this point are used as the grading benchmark to complete the risk level classification of the historical fall accident feature set, ultimately obtaining multiple graded historical fall accident feature sets. The accident features within each set have high similarity in key dimensions and can reflect the accident feature patterns under a specific risk level.

[0103] Example 5: See Figure 4After using the target spatiotemporal fusion motion feature set as the fall prevention safety monitoring result, this set is compared with a preset safety threshold set. The preset safety threshold set includes the safety range definition of various motion features, such as the normal movement speed range of personnel at different working heights, the reasonable range of body tilt angles, and the allowable range of acceleration changes. These thresholds are determined according to the safety regulations for high-altitude operations and human kinematic characteristics. During the comparison process, the deviation degree of each feature in the target spatiotemporal fusion motion feature set from the corresponding threshold in the preset safety threshold set is calculated one by one. For example, when the real-time movement speed of the personnel exceeds the normal speed threshold at the same height, the excess ratio of the speed is recorded; when the body tilt angle exceeds the reasonable range, the deviation value of the angle is calculated. By combining the deviation degrees of all features, the target motion feature deviation degree is determined through weighted calculation. This deviation degree comprehensively reflects the overall difference between the current worker's motion state and safety state.

[0104] A fall prevention warning signal is generated when the deviation of the target's motion characteristics exceeds a preset deviation threshold. The preset deviation threshold distinguishes between normal working conditions and risky conditions; exceeding this threshold indicates that the worker's motion has entered a risky zone. The fall prevention warning signal includes a warning level and a warning trigger location. Warning levels are based on the magnitude of the target's motion characteristic deviation. For example, a deviation between 1 and 1.5 times the preset threshold corresponds to Level 1; a deviation between 1.5 and 2 times corresponds to Level 2; and a deviation exceeding 2 times corresponds to Level 3. Different levels correspond to different levels of risk urgency. The warning trigger location is determined by the worker's real-time location coordinates, which are collected in real-time by positioning equipment and accurate to a specific point within the work area, such as "Latitude XX degrees XX minutes XX seconds North, Longitude XX degrees XX minutes XX seconds East, Altitude XX meters".

[0105] When transmitting fall protection warning signals to the high-altitude work monitoring terminal, the signals are first encrypted. The encryption process employs a symmetric encryption algorithm, which uses the same key for both encryption and decryption. During encryption, the content of the warning signal is converted into binary data, and the data is then permuted and replaced using the key to generate the encrypted warning signal. The encrypted signal content appears as a random sequence of characters, making it impossible to directly extract the warning information even if intercepted during transmission.

[0106] Establish a wireless communication connection with the high-altitude operation monitoring terminal. Device authentication is required before connection establishment to ensure the connected device is an authorized monitoring terminal. After successful authentication, select the communication method based on the wireless signal coverage of the work area. If a stable Wi-Fi network is available in the work area, establish a connection via Wi-Fi; if the Wi-Fi signal is unstable, switch to Bluetooth or a dedicated wireless communication frequency band to ensure the stability of the communication link. After the connection is established, monitor the signal strength in real time. When the signal strength falls below a preset value, automatically adjust communication parameters or switch the communication channel to maintain connection continuity.

[0107] Encrypted warning signals are transmitted to the high-altitude operation monitoring terminal via an established wireless communication connection. A data packet verification mechanism is used during transmission, adding a checksum to each transmitted data packet. Upon receiving the packet, the monitoring terminal verifies its integrity using the checksum. If a packet is found to be corrupted or lost, a retransmission is requested. After receiving the complete encrypted warning signal, the monitoring terminal decrypts it using a pre-stored key, restoring information such as the warning level and trigger location. Upon decryption, the terminal immediately activates an audible and visual alarm. The audible alarm uses different frequencies to distinguish warning levels; for example, a level 1 warning corresponds to a short beep every 2 seconds, a level 2 warning to a short beep every 1 second, and a level 3 warning to a continuous long beep. The visual alarm is activated by LEDs on the terminal: a yellow light for level 1, an orange light for level 2, and a red light for level 3. The flashing frequency of the LEDs increases with the warning level to ensure that on-site personnel can quickly identify the warning information and take appropriate protective measures.

[0108] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0109] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A data processing-based overhead work anti-falling safety monitoring method, characterized in that, The method comprises: obtaining an environment parameter set of a target work area, performing a matching query in a historical accident database based on the environment parameter set, and determining a historical falling accident feature set; dividing the historical falling accident feature set into a plurality of hierarchical historical falling accident feature sets according to risk levels; extracting environmental risk factors from the plurality of hierarchical historical falling accident feature sets to obtain a plurality of environmental risk factor benchmark values and a plurality of risk influence range parameters; using the plurality of risk influence range parameters as boundary conditions of a plurality of dynamic monitoring windows, capturing multi-dimensional features of real-time motion data sequences of high-altitude workers by using the plurality of dynamic monitoring windows configured, and obtaining a plurality of real-time motion feature sets; spatiotemporally correlating and fusing the plurality of real-time motion feature sets to obtain a target spatiotemporally fused motion feature set, and using the target spatiotemporally fused motion feature set as a fall-prevention safety monitoring result.

2. The data processing based overhead working anti-falling safety monitoring method according to claim 1, characterized in that, extracting environmental risk factors from the plurality of hierarchical historical falling accident feature sets to obtain a plurality of environmental risk factor benchmark values and a plurality of risk influence range parameters, comprising: extracting environmental parameters from the plurality of hierarchical historical falling accident feature sets to obtain a plurality of hierarchical historical environment parameter sequence sets; extracting critical environment parameters from the plurality of hierarchical historical environment parameter sequence sets respectively to determine a plurality of historical critical environment parameter sets, wherein each historical critical environment parameter is an environment parameter close to a falling accident occurrence threshold in each hierarchical historical environment parameter sequence; extracting factors from the plurality of historical critical environment parameter sets to obtain a plurality of historical environmental risk factor sets, and statistically analyzing the plurality of historical environmental risk factor sets to determine a plurality of environmental risk factor benchmark values; defining risk influence ranges of the plurality of hierarchical historical environment parameter sequence sets with reference to the plurality of historical environmental risk factor sets to determine a plurality of historical risk influence range sets, wherein each historical risk influence range reflects a continuous interval of environment parameters associated with a critical environment parameter in a hierarchical historical environment parameter sequence; calculating span values of the plurality of historical risk influence range sets respectively to obtain the plurality of risk influence range parameters.

3. A data processing based overhead working anti-falling safety monitoring method according to claim 2, characterized in that, extracting factors from the plurality of historical critical environment parameter sets to obtain a plurality of historical environmental risk factor sets, and statistically analyzing the plurality of historical environmental risk factor sets to determine a plurality of environmental risk factor benchmark values, comprising: extracting factors from the plurality of historical critical environment parameter sets by using a risk factor extraction algorithm to obtain the plurality of historical environmental risk factor sets; iteratively optimizing the plurality of historical environmental risk factor medians in the plurality of historical environmental risk factor sets according to a preset step size to obtain a plurality of iterated historical environmental risk factors; ​ When the dispersion coefficient of the plurality of iterative historical environmental risk factors is less than or equal to the dispersion coefficient of the plurality of historical environmental risk factors, the median of the plurality of historical environmental risk factors is taken as the plurality of environmental risk factor reference values.

4. The data processing based overhead working anti-falling safety monitoring method according to claim 3, characterized in that, Comprise: When the dispersion coefficient of the plurality of iterative historical environmental risk factors is greater than the dispersion coefficient of the plurality of historical environmental risk factors, it is judged whether the dispersion coefficient difference between the plurality of iterative historical environmental risk factors and the plurality of historical environmental risk factor medians is greater than or equal to a preset dispersion coefficient difference threshold value, if yes, the iteration is continued based on the plurality of iterative historical environmental risk factors until the maximum iteration number is met, and the plurality of iterative historical environmental risk factors obtained in the last iteration are taken as the plurality of environmental risk factor reference values; If not, stop iteration and take the plurality of iterative historical environmental risk factors as the plurality of environmental risk factor reference values.

5. A data processing based overhead working anti-falling safety monitoring method according to claim 4, characterized in that, With the plurality of historical environmental risk factor sets as a reference, the risk influence range of the plurality of hierarchical historical environmental parameter sequence sets is defined to determine a plurality of historical risk influence range sets, comprising: Using the risk factor extraction algorithm to extract environmental characteristics from the plurality of hierarchical historical environmental parameter sequence sets to obtain a plurality of hierarchical historical environmental characteristic sequence sets; A historical environmental risk factor is randomly extracted from the plurality of historical environmental risk factor sets as a first historical environmental risk factor, and a corresponding first hierarchical historical environmental characteristic sequence is matched from the plurality of hierarchical historical environmental characteristic sequence sets; According to a preset correlation threshold, interval retrieval of the first historical environmental risk factor in the first hierarchical historical environmental characteristic sequence is performed to obtain a first historical environmental risk factor interval; The time span of the first historical environmental risk factor interval is counted, and the statistical result is taken as the first historical risk influence range; According to the preset correlation threshold, the plurality of historical environmental risk factor sets are defined in the corresponding plurality of hierarchical historical environmental parameter sequence sets to determine a plurality of historical risk influence range sets.

6. A data processing based overhead working anti-falling safety monitoring method according to claim 1, characterized in that, The plurality of real-time motion characteristic sets are spatio-temporally associated and fused to obtain a target spatio-temporally fused motion characteristic set, comprising: A first real-time motion characteristic set and a second real-time motion characteristic set are randomly extracted from the plurality of real-time motion characteristic sets; The spatio-temporal correlation of the first real-time motion characteristic set and the second real-time motion characteristic set is calculated to determine a first spatio-temporal correlation set; The first spatio-temporal correlation set is standardized to obtain a first spatio-temporal correlation standardized value set; The first spatio-temporal correlation standardized value set and the second real-time motion characteristic set are subjected to matrix operation to obtain a first spatio-temporally fused motion characteristic set; A third real-time motion characteristic set is randomly extracted from the plurality of real-time motion characteristic sets, which is subjected to spatio-temporal association and fusion with the first spatio-temporally fused motion characteristic set to obtain a second spatio-temporally fused motion characteristic set; After multiple spatiotemporal correlation fusions, until the real-time motion features in the plurality of real-time motion feature sets are fused, the target spatiotemporal fusion motion feature set is obtained.

7. A data processing based overhead working anti-falling safety monitoring method according to claim 6, characterized in that, Comprise: Obtain a plurality of sample spatiotemporal correlation standardized value sets, a plurality of sample real-time motion feature sets, and a plurality of sample spatiotemporal fusion motion feature sets as training samples; Supervise and train the network model based on the recurrent neural network until the training converges, and obtain the trained spatiotemporal fusion network model; Using the spatiotemporal fusion network model, the first spatiotemporal correlation standardized value set and the second real-time motion feature set are subjected to matrix operation to obtain the first spatiotemporal fusion motion feature set.

8. The data processing based overhead working anti-falling safety monitoring method according to claim 1, characterized in that, The historical falling accident feature set is divided into risk levels to determine a plurality of hierarchical historical falling accident feature sets, comprising: Randomly extract a plurality of historical falling accident features from the historical falling accident feature set; Compare the plurality of historical falling accident features with each other to obtain a plurality of comparison combinations; Determine whether there is a comparison combination with a feature similarity less than a preset feature similarity threshold in the plurality of comparison combinations, if not, the plurality of historical falling accident features are used as a plurality of hierarchical benchmarks; Based on the plurality of hierarchical benchmarks, the historical falling accident feature set is divided into risk levels according to the preset feature similarity threshold to obtain the plurality of hierarchical historical falling accident feature sets, wherein each hierarchical historical falling accident feature set corresponds to a hierarchical benchmark.

9. The data processing based overhead working anti-falling safety monitoring method according to claim 1, characterized in that, After the target spatiotemporal fusion motion feature set is used as the anti-falling safety monitoring result, the method further comprises: Based on the target spatiotemporal fusion motion feature set and a preset safety threshold set, determine the target motion feature deviation; When the target motion feature deviation is greater than a preset deviation threshold, generate an anti-falling warning signal, the anti-falling warning signal includes a warning level and a warning trigger position; Transmit the anti-falling warning signal to the high-altitude work monitoring terminal, and the high-altitude work monitoring terminal performs sound and light alarm operation based on the anti-falling warning signal.

10. A data processing based overhead working anti-falling safety monitoring method according to claim 9, characterized in that, Transmit the anti-falling warning signal to the high-altitude work monitoring terminal, comprising: Encrypt the anti-falling warning signal to obtain an encrypted warning signal, and the encryption processing uses a symmetric encryption algorithm; Establish a wireless communication connection with the high-altitude work monitoring terminal; Send the encrypted warning signal to the high-altitude work monitoring terminal through the wireless communication connection, and the high-altitude work monitoring terminal performs sound and light alarm operation after decrypting the received encrypted warning signal.

Images