An intelligent water area virtual electronic fence based on the fusion of sound, light and electricity

Abstract

The present invention relates to the field of sensor measurement technology, and more specifically, to an intelligent water area virtual electronic fence based on acoustic, optical, and electrical fusion. It specifically comprises a water area sensor fusion network, a water area interface fluctuation feature extraction unit, an intrusion probability prediction unit, and an adaptive alarm response unit. The water area sensor fusion network is used to capture acoustic, optical, and electromagnetic signals in the target water area and dynamically attenuate and correct them based on the physical properties of the water to obtain a fused perception result. The water area interface fluctuation feature extraction unit is used to extract the fluctuation energy spectral density associated with water disturbances at various frequencies. The intrusion probability prediction unit is used to calculate the current anomaly metric and predict the intrusion probability at each location in the target water area. The adaptive alarm response unit is used to calculate the current alarm intensity at each location and execute the early warning strategy. This invention can quickly identify and amplify high-probability intrusion threats.

Description

Technical Field

[0001] The present invention belongs to the field of sensor measurement technology, and in particular relates to an intelligent water area virtual electronic fence based on acoustic, optical and electrical fusion. Background Art

[0002] In the field of water security, monitoring methods for potential risks such as illegal intrusion, floating debris threats, and poaching of fisheries have always been a key focus of the industry. Traditional water security typically relies on physical fences or mesh interceptors. While these can provide a certain degree of isolation for localized areas, they often fail to meet the requirements for accurate, all-weather, real-time monitoring in vast areas or with volatile hydrological conditions. Furthermore, many existing technologies attempt to use a single sensor type to detect water intrusions. For example, sonar detection equipment monitors underwater movements, underwater cameras or laser rangefinders perform optical identification, or low-frequency electromagnetic sensors detect suspicious metal objects. However, these single-modal solutions still face numerous challenges in practical application. First, the propagation and attenuation characteristics vary significantly in different water environments, making a single mode difficult to account for complex hydrological conditions. Second, environmental noise, bubbles, floating organisms, and hydrological fluctuations can cause a high number of false positives or false negatives, making it difficult to balance monitoring efficiency and reliability.

[0003] In existing public literature and application cases, researchers have proposed combining sonar arrays with surface infrared monitoring, hoping to leverage multi-source data to improve the ability to detect targets in water bodies. While this combination has reduced the false negative rate to some extent, it still has shortcomings in determining target distance, filtering noise, and identifying weak disturbances. Acoustic signals are easily distorted by salinity, temperature, and reflections from underwater terrain, while infrared signals are extremely sensitive to turbidity and light reflected from the water surface. Without an effective fusion algorithm, accurate target location may be lost in large-scale or multi-interference situations. Furthermore, underwater cameras, another common technology, while providing relatively intuitive image information, are often ineffective at night or in turbid waters. Furthermore, camera deployment and maintenance are costly, and the rapid attenuation of underwater light makes them uneconomical for continuous, large-scale monitoring. In contrast, electromagnetic sensing technology exhibits some advantages in detecting metal or charged targets, but electromagnetic attenuation is significant in waters with high salinity or rich mineral content, reducing the signal coverage radius or significantly reducing the ability to detect distant targets. If warnings rely solely on electromagnetic signals, it's difficult for the system to distinguish natural electromagnetic disturbances (such as geomagnetic fields and changes in seafloor rock formations) from genuine suspicious activity. Furthermore, without the ability to effectively integrate information from multiple underwater sensors, existing systems often struggle to generate an accurate global monitoring picture when encountering complex hydrological seasonal variations, the confluence of lakes and rivers, or the crisscrossing of multiple tributaries. Alternatively, they may be overly rigid in setting alarm thresholds, leading to frequent false alarms or outright misses when the target interference intensity is relatively low. Summary of the Invention

[0004] The main purpose of the present invention is to provide an intelligent water area virtual electronic fence based on the fusion of sound, light and electricity, which can greatly reduce the omissions or distortions that may be caused by a single sensing mode; through the deep fusion of multimodal data and spectrum analysis, it can effectively reduce the impact of external environmental noise on monitoring accuracy; by combining anomaly measurement with a spatiotemporal attenuation model, factors such as interference intensity, duration, and spatial distribution are incorporated into a unified probability prediction framework; and through an adaptive alarm response mechanism, it can not only quickly identify and amplify high-probability intrusion threats, but also promptly return to normal monitoring levels after the incident subsides or the disturbance is not significant, thereby maintaining high practicality and stability in changing environments, complex hydrological conditions, and diversified security scenarios, providing an effective new virtual defense line solution for various water areas.

[0005] The technical solution of the present invention is achieved as follows:

[0006] An intelligent water area virtual electronic fence based on acoustic, optical and electrical fusion, which specifically includes: a water area sensor fusion network, a water area interface fluctuation feature extraction unit, an intrusion probability prediction unit and an adaptive alarm response unit; the water area sensor fusion network is a multimodal perception network covering the target water area composed of acoustic sensors, optical sensors and electromagnetic sensors, which is used to capture the sound wave signals, optical signals and electromagnetic signals in the target water area, and perform dynamic attenuation correction based on the physical properties of water to obtain a fusion perception result; the water area interface fluctuation feature extraction unit is used to extract the fluctuation energy spectrum density related to water disturbance at various frequencies; the intrusion probability prediction unit is used to predict the intrusion probability and the adaptive alarm response unit; the water area sensor fusion network is a multimodal perception network covering the target water area composed of acoustic sensors, optical sensors and electromagnetic sensors, and ... the intrusion probability prediction unit is used to predict the intrusion probability and the adaptive alarm response unit; the water area sensor fusion network is a multimodal perception network covering the target water area composed of acoustic sensors, optical sensors and electromagnetic sensors, and the The probability prediction unit is used to set a characteristic frequency, compare the fluctuation energy spectrum density at the characteristic frequency with the template fluctuation energy spectrum density at the characteristic frequency, and compare the fusion perception result with the template fusion perception result to calculate the current anomaly measurement, set an initial detection probability, and predict the intrusion probability of each location in the target water area based on the anomaly measurement and the center position of the target water area as the nominal point; the adaptive alarm response unit is used to set a basic alarm intensity and a critical intrusion probability, calculate the current alarm intensity of each location in combination with the intrusion probability, and execute the early warning strategy based on the current alarm intensity of each location.

[0007] Furthermore, the basic alarm intensity ranges from 1 to 5, dimensionless; the current alarm intensity ranges from 1 to 10, dimensionless; the critical intrusion probability ranges from 0.7 to 0.9, dimensionless; and the initial detection probability ranges from 0.5 to 0.8, dimensionless.

[0008] Furthermore, the warning strategy includes: setting a warning threshold, if the current alarm intensity of each location is greater than or equal to the warning threshold, then issuing a warning for the location, otherwise, then not issuing a warning.

[0009] Furthermore, the fusion perception result is calculated by the following formula:

[0010] ;

[0011] in, For distance Place and time The fusion perception intensity at the time is dimensionless and ranges from 0 to 1; The initial intensity of the sound wave is in dB, ranging from 120 to 160; The initial intensity of the optical signal is in lux, ranging from 500 to 2000. is the initial strength of the electromagnetic signal, in V / m, ranging from 5 to 20; is the acoustic attenuation coefficient in water, which is related to water temperature and salinity, dimensionless, and ranges from 0.001 to 0.005; is the light signal attenuation coefficient in water, which is related to the water turbidity, dimensionless, and ranges from 0.05 to 0.2; is the electromagnetic wave attenuation coefficient in water, which is related to the water conductivity, dimensionless, and ranges from 0.01 to 0.1; The frequency of the sound wave is in Hz and ranges from 30 to 100; is the frequency of electromagnetic waves, in Hz, ranging from 1 to 10; is the refractive index of water, which is 1.33; is the speed of sound in water, in m / s, with a value range of 1450-1500; is the water density in kg / m³, with a value of 1000; is the current time, in seconds;

[0012] Furthermore, the wave energy spectral density is expressed using the following formula:

[0013] ;

[0014] in, Frequency The wave energy spectral density at , in J / Hz; is the effective height of the water surface wave, in m, with a value range of 0.01-0.1; is the acceleration due to gravity, in m / s², and its value is 9.8; is the peak frequency in Hz, ranging from 0.1 to 1; is the thermoacoustic coupling coefficient, dimensionless, ranging from 0.001 to 0.01; is the water temperature in K, ranging from 278 to 303; The conductivity of water is in S / m, ranging from 0.005 to 0.05, where S stands for Siemens; is the incident electric field strength, in V / m, ranging from 1 to 5; is the vacuum permeability, the unit is H / m, and the value is , H stands for Henry.

[0015] Furthermore, the characteristic frequency The value range is 20-50, and the unit is Hz.

[0016] Furthermore, the anomaly metric is calculated using the following formula:

[0017] ;

[0018] in, For time The abnormality measure of , dimensionless; is the characteristic frequency The wave energy spectral density at , in J / Hz; is the characteristic frequency The template fluctuation energy spectrum density at is in J / Hz and has a value range of arrive ; is the template fusion perception strength, dimensionless, ranging from 0.1 to 0.3; is the starting time, in seconds; is the characteristic development time constant, in seconds (s), with a value range of 2 to 10.

[0019] Furthermore, the invasion probability of each location in the target waters is predicted using the following formula:

[0020] ;

[0021] in, For location Location and current time The invasion probability ranges from 0 to 1; The center of the target water area; is the position uncertainty, dimensionless, ranging from 0.5 to 2; is the anomaly detection threshold, dimensionless, ranging from 1.5 to 3; is the initial detection probability; represents the Mahalanobis distance.

[0022] Furthermore, the current alarm intensity of each location is calculated using the following formula:

[0023] ;

[0024] in, For location Place and time The alarm intensity is dimensionless and ranges from 1 to 10; is the basic alarm intensity; is the critical intrusion probability.

[0025] The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity of the present invention has the following beneficial effects:

[0026] The present invention builds a multimodal perception network covering the target waters by simultaneously deploying three types of sensors: acoustic, optical, and electromagnetic. It then uniformly corrects and integrates the signal propagation, attenuation, and interaction of different modes with the water environment, thereby significantly improving the efficiency and accuracy of detecting underwater intrusions or abnormal disturbances.

[0027] First, compared to single sonar or infrared detection methods, the present invention maintains relatively stable detection capabilities in varying water qualities and temperatures, even in high turbidity and high-noise scenarios. When high turbidity causes optical sensors to fail, acoustic or electromagnetic sensing can compensate for this deficiency. And when high-salinity or high-mineral content waters severely attenuate electromagnetic waves, acoustics and optics can still perform the primary detection task. Leveraging this multimodal complementarity, the present invention significantly reduces the probability of missed detections or misjudgments, achieving an intelligent monitoring system capable of long-term operation and high sensitivity in highly variable water conditions.

[0028] Secondly, the present invention constructs a multi-angle anomaly measurement system by extracting the fluctuating energy spectrum density of water disturbances at characteristic frequencies and performing a differential comparison with historical or benchmark templates. Since the fusion perception results of sound, light, and electrical signals are integrated into the same characteristic frequency analysis framework, the system is able to accurately characterize the spectral distribution, energy peak, and temporal evolution of the disturbance signal, thereby having a stronger resolution in distinguishing daily hydrological changes from real external force intrusions. Most traditional systems can only rely on "intensity thresholds" for simple judgments, and are often prone to mistaking natural phenomena such as waves, plankton, and bubbles for intrusions. They also often turn a blind eye to low-energy real anomalies due to inappropriate threshold settings. Through fluctuating energy spectrum analysis, the present invention can effectively filter out most background clutter and can also promptly capture anomaly clues in the weak disturbance stage, greatly improving the detection stability in small targets and hydrological mixed environments.

[0029] Third, the present invention introduces an organic combination of anomaly measurement and intrusion probability prediction, and uses the energy spectrum increment at the characteristic frequency, the logarithmic ratio of the fusion perception intensity, and the time evolution factor to convert the monitored anomaly information into a probability distribution that can be distributed throughout the entire water area. This breaks through the previous limitation of "alarming only at the location of occurrence", allowing the system to continuously assess the risks of various spatial locations in the water area, and trigger an adaptive alarm response in a timely manner after the intrusion probability exceeds a certain critical value. In this way, whether the intruder sneaks in from a remote end in a slow and covert manner, or causes strong interference in a local area in a short period of time, the system can track and identify it through time attenuation or space attenuation models, and will not be passive or lagging in a few monitoring blind spots or where the sensor sensitivity is insufficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1A schematic diagram of the system structure of a well inclination and geomagnetic field strength weight measurement system based on a high-precision fluxgate sensor provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0031] In the following description, specific details such as specific system structures, interfaces, and technologies are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present invention. However, it will be apparent to those skilled in the art that the embodiments of the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known devices, circuits, and methods are omitted to avoid obscuring the description of the embodiments of the present invention with unnecessary detail.

[0032] Example 1, reference Figure 1 :An intelligent water area virtual electronic fence based on acoustic, optical and electrical fusion, which specifically includes: a water area sensor fusion network, a water area interface fluctuation feature extraction unit, an intrusion probability prediction unit and an adaptive alarm response unit; the water area sensor fusion network is a multimodal perception network covering the target water area composed of acoustic sensors, optical sensors and electromagnetic sensors, which is used to capture the sound wave signals, optical signals and electromagnetic signals in the target water area, and perform dynamic attenuation correction according to the physical properties of water to obtain a fusion perception result; the water area interface fluctuation feature extraction unit is used to extract the fluctuation energy spectrum density related to the water body disturbance at each frequency; the intrusion probability prediction unit is used to extract the fluctuation energy spectrum density related to the water body disturbance at each frequency; the intrusion probability prediction unit is used to detect the intrusion probability of the water area ... The probability prediction unit is used to set a characteristic frequency, compare the fluctuation energy spectrum density at the characteristic frequency with the template fluctuation energy spectrum density at the characteristic frequency, and compare the fusion perception result with the template fusion perception result to calculate the current anomaly measurement, set an initial detection probability, and predict the intrusion probability of each location in the target water area based on the anomaly measurement and the center position of the target water area as the nominal point; the adaptive alarm response unit is used to set a basic alarm intensity and a critical intrusion probability, calculate the current alarm intensity of each location in combination with the intrusion probability, and execute the early warning strategy based on the current alarm intensity of each location.

[0033] Specifically, during operation, acoustic sensors primarily capture the propagation characteristics of underwater sound waves, identifying underwater activity and disturbances by analyzing sound intensity, phase changes, and reflection patterns. Optical sensors exploit the transmission and scattering of visible or infrared light in water to monitor signal attenuation and spectral shifts caused by turbidity, plankton, or other impurities. Electromagnetic sensors are sensitive to the attenuation of low-frequency electromagnetic fields in water, as well as field intensity anomalies caused by charged objects or conductive media. To enable the processing of these three heterogeneous signals under the same standard, the system incorporates real-time measurement methods for environmental parameters such as water temperature, salinity, turbidity, and conductivity, and incorporates these into a dynamic attenuation correction module, thereby adaptively compensating for the energy propagation path of each signal. This fusion network not only achieves synchronized data acquisition and buffering for each sensor node at the hardware level, but also, at the software level, uses a multi-source data fusion algorithm to combine the intensity characteristics of the acoustic signal, the transmission characteristics of the optical signal, and the field strength characteristics of the electromagnetic signal to produce a unified, integrated perception metric. This fusion process weights each sensor's noise level, measurement accuracy, and transient changes in the water environment, ensuring accurate extraction of true environmental disturbance information even in multi-interference scenarios. When the sensor network detects abnormal fluctuations in the water, the system transmits the fused perception results to subsequent units and performs depth comparisons based on a priori fluctuation templates. Because the fusion network's construction fully accounts for the nonlinear absorption and scattering properties of water, the output data not only preserves the sensitivity characteristics of the acoustic, optical, and electromagnetic signals, but also effectively suppresses single-mode distortion caused by factors such as water quality, environmental clutter, and bubbles. Through this comprehensive information acquisition and correction approach, the water sensor fusion network lays a highly reliable foundation for subsequent fluctuation feature extraction, intrusion probability prediction, and alarm response. It truly achieves the precise conversion of raw physical water signals into comprehensive judgment indicators, enabling the system to maintain high response sensitivity and stability in complex underwater environments, thereby enhancing the applicability and practical value of this intelligent water virtual electronic fence for security monitoring and real-time early warning.

[0034] The water interface fluctuation feature extraction unit is a key step in the present invention's deep dive into detailed water surface and water body disturbances based on multimodal sensing data. It utilizes a combined analysis of acoustic, optical, and electromagnetic signals to accurately determine the dynamic energy distribution of the target water area, providing a more robust and sensitive basis for determining the presence of abnormal intrusions. In operation, this unit preprocesses the multimodal signals transmitted from the water sensor fusion network to minimize interference from background noise and random environmental fluctuations. It then distinguishes common hydrological disturbances (such as small surges, bubble formation, and plankton activity) from high-energy disturbances likely caused by intrusion or external interference based on the degree of characteristic coupling across different frequency ranges. Next, this unit comprehensively considers the modulation effects of acoustic wave propagation on underwater sound waves due to factors such as temperature, salinity, and pressure. It also incorporates the refraction and scattering of optical signals in water, compares the sensitivity of electromagnetic sensing data to charged or metallic targets, and uses frequency domain or time-frequency transformation techniques to extract the energy spectrum distribution directly related to water interface fluctuations. By mapping the energy contribution of multimodal signals into a unified frequency domain or feature domain, this unit can fully characterize the evolution of underwater disturbances in the same time window, and identify key disturbance sources by detecting energy peaks and bandwidths in different frequency bands. When the system captures sudden high-energy abnormal changes, this unit will extract the fluctuation energy spectral density characteristics at this time, focusing on analyzing its asymmetry, excessive sharpness, or degree of deviation from historical template data in the frequency domain, and verifying whether the disturbance is consistent with known natural factors within the water area through a backtracking algorithm when necessary. Once an energy concentration pattern that is completely different from common hydrological characteristics is found, this unit will mark it as a potential abnormal signal and compare it with the template feature value for difference, thereby obtaining a preliminary quantitative assessment of the disturbance. Subsequently, this unit sends the calculated spectral feature parameters or energy difference information to the intrusion probability prediction unit, providing the necessary input basis for subsequent abnormal measurement judgment and spatial distribution prediction. Because the unit combines acoustic, optical, and electromagnetic signals for multi-angle perception of the water interface and underwater environment during its construction, and distinguishes between normal and abnormal disturbance patterns through in-depth analysis of frequency domain energy distribution, it can maintain high detection sensitivity in low-energy disturbance environments while reducing false alarms and missed alarms under high-energy interference conditions. At the same time, this method of extracting fluctuation characteristics can accurately identify low-intensity disturbances associated with biological activity or environmental changes, further enhancing the applicability of intelligent water virtual electronic fences based on acoustic, optical, and electrical fusion in complex underwater environments. This allows them to effectively measure potential risk factors in both high-wave conditions and relatively calm waters, playing a crucial supporting role in subsequent intrusion probability prediction and adaptive adjustment of alarm intensity.

[0035] The intrusion probability prediction unit is a key component of the present invention's dynamic determination and spatial distribution assessment. It comprehensively analyzes the energy spectrum information output by the water interface fluctuation feature extraction unit and the fused perception results to derive the intrusion risk probability at different locations. In actual operation, the unit first evaluates the difference between the current spectral energy and historical template data based on pre-set characteristic frequencies. If a significant deviation between the characteristic signal within the water area and the template is detected, the system calculates a corresponding anomaly metric to reflect the strength and persistence of the disturbance. To extend this anomaly metric to the entire water area, the unit uses the center of the water area or a predefined reference coordinate as a reference point. Drawing on understanding the propagation characteristics of acoustic, optical, and electromagnetic signals in water, it introduces specific spatial and temporal attenuation models, combined with a priori initial detection probabilities, to estimate the potential risk value at each location. This design logic is based on the understanding of underwater disturbance propagation: most intrusions or external interventions generate disturbance waves in the surrounding water. These disturbances gradually decay with distance and may exhibit different energy decay curves due to variations in factors such as water temperature, salinity, and conductivity. Based on this, the intrusion probability prediction unit assigns weights to the fused perception intensity and corresponding fluctuation energy spectrum deviation at different distances during its calculations to distinguish the different risk levels posed by close-range high-energy disturbances from those posed by more distant, weak disturbances. In this process, the initial detection probability is set as a priori, reflecting the system's understanding of the current water state. If the water area has a history of repeated intrusions or disturbances, the initial detection probability can be set higher in the next round of detections to increase sensitivity to suspicious activity. Conversely, if the water environment is relatively stable, the initial detection probability can be appropriately lowered to reduce false alarms caused by natural fluctuations. Furthermore, to respond to potential short-term, drastic changes, the unit incorporates temporal evolution into its intrusion probability calculations: a rapid increase in the anomaly metric over a short period of time indicates the occurrence of a sudden, strong disturbance; a stable or slowly increasing anomaly over a longer period of time suggests a more gradual change in the hydrological environment itself. By combining temporal and spatial factors, the intrusion probability prediction unit provides a global assessment of potential risks at every coordinate within the water area and at any given moment, providing distributed and dynamically evolving input for the subsequent adaptive alarm response unit. Once the system identifies a location where the risk exceeds an acceptable threshold based on the probability output by the prediction unit, the subsequent alarm mechanism rapidly adjusts the intensity of the corresponding acoustic and optical signals or triggers coordinated early warning devices, thereby responding to the potential security threat with minimal delay.It is worth noting that in order to maintain computational stability and accuracy under different environmental conditions, the intrusion probability prediction unit will recalibrate core parameters during initialization or periodic maintenance, such as re-measuring the salinity, temperature, or turbidity value of the water area, and correcting the initial detection probability and distance attenuation coefficient based on the most recent alarm record, so that the intrusion probability estimation is closer to the actual underwater propagation law. Through the above process, the intrusion probability prediction unit plays a key role in the intelligent water virtual electronic fence system proposed by the present invention. It can not only effectively inherit the rich information from the fluctuation feature extraction unit, but also cooperate with subsequent alarm responses to form an adaptive and evolvable intrusion judgment closed loop, allowing the system to maintain high-precision positioning and timely capture of potential risks in the face of various sudden disturbance scenarios, thereby providing a stronger level of reliability and intelligence for water safety protection.

[0036] The adaptive alarm response unit is a key module in the present invention, enabling rapid and targeted policy adjustments upon detecting potential intrusion signs. By meticulously processing the spatial probability distribution results from the intrusion probability prediction unit, it dynamically analyzes the risk level of each monitored location and, based on the system's baseline alarm intensity and critical intrusion probability, dynamically schedules responses at different times. In operation, the unit first reads the real-time intrusion probability value for each location and then sets an initial warning output plan based on the baseline alarm intensity, ensuring that any anomalies are promptly addressed in the relevant areas. Furthermore, the adaptive mechanism proposed in this invention allows the system to adopt differentiated alarm levels for different risk levels: when the risk at a certain location is still low, the alarm level is maintained at the lowest level to avoid excessive disruption to the public or waste of system resources. However, when the risk rapidly increases and exceeds the critical intrusion probability, the alarm level is automatically raised to a higher level, triggering more prominent audio and visual cues or further external warning measures. This allows the system to balance comprehensive monitoring over a large area with focused protection of localized areas. Furthermore, the adaptive alarm response unit takes into account the decaying nature of time. When the intrusion probability at a location increases significantly over a short period of time, the system determines that the location is more likely to face sudden intervention. Therefore, it increases the alarm intensity based on the real-time anomaly metric, alerting surrounding monitoring and management personnel to prepare for emergency intervention or investigation. If the anomaly is only a brief peak that quickly returns to a safe range, the alarm intensity will also decay over time to avoid unnecessary high-intensity alerts. This adaptive scheduling approach ensures that the alarm response at each location throughout the monitoring cycle closely matches its actual risk status, reducing false alarms caused by natural fluctuations in the water environment while also promptly detecting potential intrusions that might otherwise be overlooked. More importantly, this adaptive alarm strategy provides system operators with the ability to implement tiered management. For example, in key areas or high-risk boundaries, the base alarm intensity can be artificially increased, or the critical intrusion probability threshold can be lowered, thereby triggering higher-level monitoring and warnings in sensitive waters in advance. On the other hand, in areas with a more stable background environment or low surrounding risk, more moderate thresholds and alarm intensities can be used, ensuring safety while minimizing disruption to nearby ecosystems or normal activities.It is worth mentioning that in order to enable the adaptive alarm response unit to operate effectively under different hydrological conditions, the present invention deeply couples the fusion perception results of acoustic, optical, and electromagnetic signals with the unit: in high turbidity or low light environments, the calculation method of the basic alarm intensity will be dynamically modified according to the sensor sensitivity to avoid the recognition inaccuracy that may result from relying on a single mode; in heterogeneous waters, different depths, or complex hydrological seasonal changes, the critical intrusion probability will also be fine-tuned based on the phased calibration data to ensure that any new abnormal disturbances can be captured in time and trigger the corresponding alarm dispatch. Through this multi-dimensional and adaptive approach, the present invention has greatly improved the accuracy and real-time performance of alarms. In addition, in the face of persistent interference and random short-term disturbances, it can output different levels of acoustic, optical, and electrical feedback in a targeted manner, allowing managers to judge the severity of potential threats based on the alarm level and make reasonable deployments.

[0037] Example 2: The basic alarm intensity ranges from 1 to 5, dimensionless; the current alarm intensity ranges from 1 to 10, dimensionless; the critical intrusion probability ranges from 0.7 to 0.9, dimensionless; the initial detection probability ranges from 0.5 to 0.8, dimensionless.

[0038] Specifically, in the intelligent water virtual electronic fence system constructed by the present invention, the basic alarm intensity ranges from 1 to 5. This dimensionless characterization method organically connects the objective environmental differences of different water areas with the system's subjective warning requirements, facilitating unified comparisons across diverse scenarios and ensuring rapid optimization through corresponding range mapping when environmental conditions or monitoring requirements change. The upper limit is defined within 5 because, through extensive experiments and historical data analysis, it has been found that basic alarm intensities exceeding this value in practical applications result in excessive energy consumption and operational burdens on warning equipment, making them unsuitable for long-term, continuous monitoring. Setting the lower limit too low fails to fully reflect the basic defense intent against the target water area, thereby weakening the system's initial vigilance. The current alarm intensity ranges from 1 to 10, established based on a comprehensive consideration of the potential for multiplication of the basic alarm intensity and the flexibility of security management. By allowing the system to gradually increase the intensity from low to high levels during subsequent adaptive calculations, it can better match the various forms and severity of disturbances. When a localized area encounters a rapid and high-risk intrusion signal, the current alert intensity will approach the upper limit of 10. However, in the face of minor disturbances or less suspicious situations, the alert intensity can be maintained slightly above the base alert intensity, thus achieving effective protection while balancing energy conservation and disturbance reduction. The critical intrusion probability range (0.7 to 0.9) is primarily determined by the confidence and plausibility of identifying potential intrusions after multimodal data fusion. When the intrusion probability exceeds 0.7, the system begins to consider the location as a "high-risk area" and appropriately raises the alert level. When the probability approaches or exceeds 0.9, it indicates a high probability of intrusion or abnormal activity in the area, necessitating the rapid initiation of a higher-level alert or emergency response. Setting the critical intrusion probability too low can easily lead to excessive false alarms, disrupting the smooth operation of the system. However, setting it too high can lead to missed alarms or a delay in responding to real intrusions. The range of the initial detection probability (0.5 to 0.8) reflects the system's prior assessment of the historical safety status of the waters and monitoring needs. Its lower limit, generally no lower than 0.5, maintains the system's basic sensitivity. After all, in completely unknown or risky situations, a certain level of basic detection capability is necessary. The upper limit of 0.8 is based on engineering experience and field analysis: if the initial detection probability is higher than this value, the system's misjudgment rate for short-term, weak disturbances will increase significantly, while also increasing the unnecessary weighting of alarm intensity in the early stages. By adjusting within this range based on actual water conditions, the system can adaptively adjust its initial response to suspicious activity.It should be pointed out that the values of these parameters will also be affected by factors such as water area, sensor density, water quality conditions and seasonal climate in specific deployment. For example, in large marine environments, if the water temperature and salinity changes significantly, the system will tend to choose a higher basic alarm intensity and critical intrusion probability to improve the overall prevention level; on the contrary, in small waters that are relatively closed and have no obvious ecological threats, the initial detection probability can usually be moderately reduced to reduce the system sensitivity and reduce the waste of resources and unnecessary warning interference caused by overreaction.

[0039] Example 3: The warning strategy includes: setting a warning threshold, if the current alarm intensity of each location is greater than or equal to the warning threshold, then issuing a warning for the location, otherwise, no warning is issued.

[0040] Specifically, in the intelligent water area virtual electronic fence system of the present invention, to ensure comprehensive monitoring coverage while implementing a more effective localized early warning strategy, Example 3 proposes a determination scheme based on the comparison of current alarm intensity with an early warning threshold. Specifically, after receiving the prediction results for each location from the intrusion probability prediction unit, the system combines the current alarm intensity calculated by the adaptive alarm response unit to provide real-time updates on the spatial security status of the water area. When the current alarm intensity at a monitored location is greater than or equal to the set early warning threshold, the system marks the location as a high-risk or highly likely intrusion target and then issues multimodal warning feedback, such as acoustic, optical, or electromagnetic, for this target. Furthermore, the system can also trigger external alarms, such as notifying patrol boats or monitoring centers, to enable rapid location and intervention. Conversely, if the current alarm intensity remains below the early warning threshold, the abnormality at the location is insufficient to be considered a substantial threat, and the system will maintain basic, low-level monitoring without consuming further warning resources. In this way, this early warning strategy, centered on "current alarm intensity versus warning threshold," can not only trigger timely alarms or reinforcements when disturbances are intense, but also reduce false alarms and energy consumption when disturbances are low or unconfirmed, thereby enabling optimal configuration of water safety monitoring at a refined spatial scale. During its implementation, this strategy can also intelligently adjust the warning threshold based on historical data and environmental characteristics: if a region has experienced multiple real intrusion incidents, the system can lower the threshold to increase alarm sensitivity; conversely, for areas with relatively stable disturbances and a higher security level, the threshold can be appropriately raised to avoid excessive emergency response, thereby balancing actual protection needs with the ecological balance of the waters.

[0041] Example 4: The fusion perception result is calculated using the following formula:

[0042] ;

[0043] in, For distance Place and time The fusion perception intensity at the time is dimensionless and ranges from 0 to 1; The initial intensity of the sound wave is in dB, ranging from 120 to 160; The initial intensity of the optical signal is in lux, ranging from 500 to 2000. is the initial strength of the electromagnetic signal, in V / m, ranging from 5 to 20; is the acoustic attenuation coefficient in water, which is related to water temperature and salinity, dimensionless, and ranges from 0.001 to 0.005; is the light signal attenuation coefficient in water, which is related to the water turbidity, dimensionless, and ranges from 0.05 to 0.2; is the electromagnetic wave attenuation coefficient in water, which is related to the water conductivity, dimensionless, and ranges from 0.01 to 0.1; The frequency of the sound wave is in Hz and ranges from 30 to 100; is the frequency of electromagnetic waves, in Hz, ranging from 1 to 10; is the refractive index of water, which is 1.33; is the speed of sound in water, in m / s, with a value range of 1450-1500; is the water density in kg / m³, with a value of 1000; The current time, in seconds.

[0044] Specifically, The value ranges from 0 to 1 and is used to quantify the distance between the target waters and the target area. and time The total energy response under these conditions. Considering it as a weighted and dynamically corrected intensity index will help to make sensitive judgments on possible intrusion activities. Among the three additional terms in this formula, the first term mainly reflects the attenuation process of underwater sound waves as they evolve with distance and time. It uses the initial intensity of the sound waves to calculate the attenuation process of underwater sound waves. To characterize the relative energy level of the signal at the time of transmission, and to introduce an exponential attenuation factor due to water absorption and scattering during propagation ,in is the acoustic attenuation coefficient in water, which is generally between 0.001 and 0.005. This numerical range is the result of a large number of hydroacoustic studies and engineering tests, and can cover the average absorption of sound waves in fresh water, brackish water, and slightly saline environments. On the one hand, the sound wave frequency Incorporating this into consideration, considering that when the sound wave operates in the range of 30 to 100 Hz, its energy distribution and detectable range underwater are different, and on the other hand, the time The square root of also realizes a diffusion-type attenuation correction: the closer to the sound source emission moment, the more concentrated the energy; over time, the sound energy gradually diffuses and decays, which is consistent with the law of sound wave distribution in actual water environment. The second term describes the process of optical signal penetration in water, where the initial light intensity The range of 500 to 2000 lux takes into account the light intensity required for monitoring environments in different waters. For example, in turbid or deep water environments, a higher initial light intensity is required to maintain a certain detection capability, while in relatively clear shallow water areas, a lower power light source may be used. It is used to reflect the impact of water turbidity, suspended matter concentration and other factors on the propagation of optical signals. It usually fluctuates between 0.05 and 0.2. If the water body itself contains rich algae or sediment, will be larger, and the light will be significantly attenuated in a shorter distance; on the contrary, in clearer waters, At the same time, light follows a certain geometric diffusion law in the medium, so the denominator appears ,here The value is 1.33, which represents the refractive index of water. It is used to correct the refraction deviation and scattering loss that occurs when light propagates at the water-medium interface, ensuring that the formula can maintain good applicability in different refractive index environments. It is the classic point light source radiation diffusion model, which represents the universal law that the luminous flux decreases with the square of the distance. The third term focuses on the propagation characteristics of electromagnetic waves in water, through the initial electromagnetic intensity It is coupled with a complex attenuation factor to simulate the absorption characteristics of water to electromagnetic fields. The range of 0.01 to 0.1 is to accommodate different conductivity levels from low-salinity freshwater to high-salinity seawater. In a water environment with high salinity and strong conductivity, electromagnetic waves will be rapidly attenuated, while in areas with low mineral content or pure water, the attenuation rate will be relatively slow. When the frequency varies between 1 and 10 Hz, the signal's ability to penetrate water is still limited. However, the lower the wavelength (i.e., the smaller the frequency), the slower the attenuation will be. Therefore, the introduction of It is used to reflect the difference in loss of electromagnetic waves of different frequencies in water, which is consistent with the attenuation trend of underwater electromagnetic fields obtained in actual tests. In order to make the energy synthesis of the three major signals finally mapped to the same measurement scale, this formula introduces The factor is used to balance the differences in medium properties involved in underwater sound propagation and electromagnetic attenuation. , the present invention is set to 1000 , which is basically a reference value for the density of fresh water, but it can also be approximately regarded as this level in a slightly saline environment; and the speed of sound Generally between 1450 and 1500 It is used to reflect the effect of water temperature change or water depth difference on sound wave transmission. The speed of sound will fluctuate slightly under different temperature or salinity conditions. Therefore, during the actual system operation, this parameter can be corrected through periodic hydrological testing. Increase and time As time goes by, these three items gradually decrease according to their independent physical attenuation mechanisms. However, in some special scenarios, the acoustic signal may be very strong while the optical signal and electromagnetic signal may be very weak, or the environment may be very clear and the contribution of the optical signal may be very obvious. In this paper, the three items are arithmetically superimposed to obtain a total fusion strength value. , and forcibly normalize it to the range of 0 to 1, so that the subsequent water disturbance detection algorithm and intrusion probability prediction module can compare the signal levels at different locations and times under a unified indicator system. When it is close to 1, it often means that there is a high energy distribution at the monitoring point, which may be caused by a large sound source or strong electromagnetic interference in the surrounding area, or the local light reflection is particularly obvious, which is worthy of further determination whether it is an intruder or an abnormal target; on the contrary, when When it approaches 0, it means that all three signals have been greatly attenuated, the water area is extremely calm, and it can be basically determined that there is no large-scale abnormal activity. Behind this hierarchical modeling and parameter selection, it also reflects the design idea of the present invention in terms of dynamic optimization of compatible multiple environmental parameters: before deployment or during operation, the system will select the appropriate 、 、 、 、 When the water environment undergoes seasonal or event-related changes, these attenuation coefficients are recalibrated through the back-end algorithm to ensure that the fusion perception results accurately reflect the actual water state. and electromagnetic initial intensity You can also make appropriate choices based on security protection needs and energy supply capabilities. If a certain area of water is small and has sufficient light around it, the system does not need to invest in too high a light intensity. If there are a large number of conductive impurities in the water environment and the interference source is unpredictable, the initial electromagnetic intensity can be slightly increased to enhance the detection sensitivity of strong interference situations. Through such flexible parameter configuration and the coordination of multiple signal channels, the present invention constructs a complete quantitative link from physical energy to normalized intensity, so that the subsequent water interface fluctuation feature extraction and intrusion probability prediction do not need to switch repeatedly between different dimensions, and it is also easier to combine historical template data to determine whether the current fluctuation characteristics are within a safe range or there is a large deviation. Ultimately, this fusion perception result This will run through the subsequent anomaly metric calculation and intrusion probability space mapping process to ensure that the system can continue to provide high-confidence monitoring information in complex underwater environments, allowing the smart water virtual electronic fence to quickly identify various potential threats in large-scale deployment and dynamic tracking, and cooperate with the adaptive alarm response unit to jointly maintain water safety.

[0045] Example 5: The wave energy spectrum density is expressed using the following formula:

[0046] ;

[0047] in, Frequency The wave energy spectral density at , in J / Hz; is the effective height of the water surface wave, in m, with a value range of 0.01-0.1; is the acceleration due to gravity, in m / s², and its value is 9.8; is the peak frequency in Hz, ranging from 0.1 to 1; is the thermoacoustic coupling coefficient, dimensionless, ranging from 0.001 to 0.01; is the water temperature in K, ranging from 278 to 303; The conductivity of water is in S / m, ranging from 0.005 to 0.05, where S stands for Siemens; is the incident electric field strength, in V / m, ranging from 1 to 5; is the vacuum permeability, the unit is H / m, and the value is , H stands for Henry.

[0048] Specifically, the core idea is to integrate the classical spectrum function of water surface waves, thermoacoustic coupling terms, and electromagnetic coupling terms, so as to more realistically depict the energy distribution at the interface and inside the water body under the multimodal monitoring system. First, from a physical point of view, the first term in the formula Derived from the typical water surface wave spectrum description, it is usually used to reflect the Wave energy varies with effective wave height The changing situation, The value range is 0.01 to 0.1 meters, covering everything from calmer swells to relatively significant surface waves. is the basic external driving force. When the peak increases, the energy in the corresponding frequency band will also increase. This shows that the waves exhibit typical high-frequency rapid attenuation in the frequency domain, which is consistent with the empirical and theoretical spectral distribution characteristics of deep-water gravity waves. The peak frequency in (0.1 to 1 Hz) is used to locate the main energy concentration area of the spectrum. Under actual hydrological conditions, if the waves are relatively stable, Often tend to the low end; if the waters encounter external disturbances such as wind and waves, It will move up or down, directly affecting the distribution shape of this part of the spectrum.

[0049] Secondly, the second item It represents the contribution of thermal-acoustic coupling to the total energy in the frequency domain, reflecting the additional effects of water temperature, sound velocity, medium density and other factors on water disturbance in a relatively high frequency band. When rising, the thermal motion of microscopic particles in the medium is enhanced, which leads to changes in the way the medium transmits and absorbs sound energy, and the speed of sound increases. Water density It also has a modulation effect on the propagation of the sound field, either amplifying or weakening it. It is a dimensionless thermal-acoustic coupling coefficient, and its value range is generally between 0.001 and 0.01. If there is a large temperature gradient or heat source in the water area, the coefficient will increase relatively, making the modulation of the overall energy spectrum by water temperature fluctuation more obvious; if the temperature distribution in the water area is relatively uniform, can be kept at a low level, making the contribution relatively limited. The disturbance under the action of electromagnetic waves is mainly taken into consideration, which corresponds to the coupling effect between electromagnetic excitation and water conductivity observed during underwater electromagnetic sensing in the present invention. is the electrical conductivity of water in Siemens per meter ( ), fluctuates between 0.005 and 0.05, and usually takes a larger value in an environment with higher salinity; at the same time, the incident electric field intensity On 1 to 5 Between, it indicates the field strength applied during sensing or detection; The frequency of the electromagnetic signal (1 to 10 Hz). If the frequency is too high, the attenuation in the water will be very obvious and cannot be reflected in the main energy distribution of this item. If the frequency is too low, the electromagnetic wave can penetrate the water more deeply and significantly change the energy spectrum of the local water area. Pick , which is the basic constant for the calculation of all electromagnetic phenomena. It is also used here to normalize the energy conversion of electromagnetic disturbances so that it can still be reflected on a unified energy scale after adding it to the sum of water waves and thermoacoustic coupling.

[0050] It should be emphasized that the above three items cannot simply be linearly superimposed to fully describe all disturbances in the water, but they have effectively covered the three main energy sources of water surface gravity waves, thermoacoustic fluctuations and electromagnetic interactions. When the system monitors and collects real multimodal data, it will confirm it through data fitting or parameter backtracking. 、 and The specific values of key variables such as , make this formula still have sufficient applicability when dealing with specific environments in different waters. For example, in open waters significantly affected by wind and waves, the first term (wave spectrum) often dominates, and the system will observe significant Attenuation and strong energy distribution caused by the nearby peak frequency; in tropical or geothermal waters with obvious water temperature gradient, the contribution of the second term increases rapidly, forming a mixed spectrum with a considerable proportion with the first term; for example, in situations with high salinity or underwater electromagnetic interference, the third term It will play a more critical indicative role. In general, this formula gives the water interface fluctuation characteristics a broader and more multi-dimensional descriptive capability under the system of the present invention. Through the "spectral level" analysis of the three main disturbance mechanisms, the system can capture abnormal changes beyond the normal fluctuation range, providing a more reliable basis for subsequent abnormal measurement and intrusion probability prediction. If a spectrum peak significantly higher than the normal level appears in a certain frequency band, it often means that an additional disturbance source has occurred here (possibly an intruder or other foreign object), thereby triggering the depth comparison and alarm mechanism of the subsequent modules of the present invention. It can be said that through the clever combination and parameter calibration of such a set of multi-source physical factors, the complex disturbances in the water area can be analyzed in the frequency domain as the sum of three main energy components, allowing the intelligent water area virtual electronic fence to achieve refined identification and sensitive tracking of fluctuations under the multimodal perception framework.

[0051] Example 6: Characteristic Frequency The value range is 20-50, and the unit is Hz.

[0052] Specifically, in the intelligent water area virtual electronic fence system involved in the present invention, the characteristic frequency The value range of is set between 20Hz and 50Hz, primarily based on a combination of empirical and theoretical results from multiple field measurements and spectral analysis of underwater disturbance signals. First, from the perspective of underwater acoustics, although ultra-low-frequency signals below 20Hz have stronger penetration and longer propagation distances, they are often associated with large-scale geological or climatic disturbances and are not suitable for fine-grained identification of localized intrusions. Frequency ranges above 50Hz are typically at the lower end of the near-ultrasonic range and can be significantly affected by noise and signal attenuation in aquatic environments. Their effective detection radius is also significantly reduced, making it difficult for the system to balance coverage efficiency and sensitivity when monitoring large areas of water. Second, from a practical monitoring perspective, selecting 20Hz to 50Hz as the characteristic frequency range can capture the most representative low- and medium-frequency disturbance signatures in most freshwater or nearshore environments. These disturbance signals are often caused by aquatic biological activity, small intrusive objects, or other external forces. They have relatively distinct energy peaks in the spectrum, distinguishing them from natural surface waves, background noise, and even high-power acoustic waves generated by ships. Third, from the perspective of algorithm design and hardware implementation, data sampling and real-time processing within the 20Hz to 50Hz range are relatively easy to implement: ordinary underwater acoustic sensors or underwater microphones can accurately capture signals within this bandwidth, and preliminary data cleaning and feature extraction can be completed without the need for high sampling rates or complex filtering circuits. At the same time, once an energy peak significantly above the normal background level is detected within this frequency range, subsequent anomaly measurement formulas can be used to more sensitively analyze its spatial and temporal distribution, distinguishing between various possible sources of disturbance (including unintentional natural disturbances and potential illegal intrusions). Finally, the present invention found in long-term testing that if the characteristic frequency is selected below 20Hz or above 50Hz, it will not only increase the high hardware cost, but also lead to a decrease in the detection efficiency of common intrusion scenarios; while locking it within this relatively concentrated and adjustable range, the characteristic frequency can be adapted to most water environments, covering low-noise areas such as freshwater lakes, rivers, and reservoirs, and can also cope with medium waves and biological activities in near-coastal zones or semi-enclosed bays, better balancing the effective working range of multimodal perception and the actual deployment difficulty, so that the present invention can still achieve high detection accuracy and stability under different hydrological conditions, providing a reliable data basis for subsequent intrusion probability prediction and alarm strategy formulation.

[0053] Example 7: The anomaly metric is calculated using the following formula:

[0054] ;

[0055] Among them, this definition unifies the time-frequency information and historical reference values in the water area from three dimensions to characterize the severity of the intrusion or abnormal disturbance in multiple ways. Specifically, the first dimension reflects , that is, at the characteristic frequency The observed wave energy spectral density Compared with the reference template value in the historical or normal state The difference ratio. and The units are , take the difference between the two and divide by After that, a dimensionless ratio of relative increment can be obtained. The larger the value, the higher the energy of the current water area at the characteristic frequency is, which is likely caused by additional disturbance or high-energy excitation caused by intrusion. If the value is close to zero or negative, it means that the energy of the water area in this frequency band is basically similar to the template or lower than the normal level, and the system can be considered that no significant abnormality has occurred. The value range is limited to arrive , is an empirical range derived from a large amount of hydrological and wave observation data: Generally speaking, such an energy spectrum density range can cover both still water or slight fluctuations, and can also adapt to the characteristic frequency bandwidth brought by moderate winds and waves.

[0056] The second dimension is , which makes a logarithmic ratio of the three-modal fusion perception strength, and further converts the real-time measured Template strength For comparison. The sum and attenuation characteristics of the three signals of sound, light and electromagnetic can best reflect the comprehensive energy level of multi-source disturbance in the same time window; when compared with its template value When the ratio is calculated and the logarithm is taken, if the result is greater than 0, it means that the current fusion strength exceeds the normal or reference level, and the larger the amplitude, the more obvious the difference; if the result is less than 0, it means that the current perception strength is lower than the template, which can help the system determine whether it is a simple natural fluctuation. Template The range of 0.1 to 0.3 is typically selected based on historical averages in deployment scenarios or baseline strength under small disturbances; its upper limit is controlled at 0.3 to ensure that, in most cases, the fusion strength during normal activity does not increase unchecked, rendering the logarithmic ratio meaningless. This logarithmic operation can further amplify or reduce the difference in the ratio between the two, enabling sensitive detection of overshooting behavior even in weak signals, while also preventing excessively large ratios from causing imbalanced anomaly metrics in strong signals.

[0057] The third dimension is the time correction factor , which incorporates the temporal evolution of abnormal disturbances in the water into the measurement, allowing the system to have a higher abnormality score for those "persistent and strong" disturbances, and can also more gently increase the abnormality value when there are only short instantaneous spikes or random small disturbances. Specifically, Represents the start time, which is often used to mark the start point of the current detection cycle or event observation; The value range of is set to 2 to 10 seconds (or longer), which is consistent with the diffusion or decay period of acoustic, optical and electromagnetic disturbances in most waters. For example, if a sudden disturbance appears and disappears quickly in a very short time, It won't be big enough to The factor is close to 1, thus avoiding giving too high anomaly scores to short-term peaks; if the disturbance lasts for a long time or even does not decay significantly within a few seconds to more than ten seconds, the factor gradually approaches 1, reflecting that the disturbance has become a situation worthy of high vigilance.

[0058] By multiplying these three physical dimensions and mathematical forms, the system can capture more accurate and spatially resolved anomaly measurements in the water. .if is much greater than 1, it means that significant additional energy release or interference occurs near the characteristic frequency; at the same time, if If it is positive and the value is large, it also means that the fusion perception strength significantly exceeds the normal value; if Also approaches 1, indicating that the interference lasts for a long time, so after multiplying the three, It will reach a high value, which will trigger the system to mark this abnormality as dangerous or requiring high attention in the subsequent intrusion probability prediction and adaptive alarm stages. In the opposite case, if there is no obvious spectrum increment in the characteristic frequency bandwidth, or the fusion perception result is not much different from the template, and the interference is short-term transient, then It may even be close to 0; or Less than 0; or the time correction factor has not risen enough to show a significant impact, then the final The system will regard this situation as a tolerable random disturbance or normal fluctuation and will not further raise the alarm level.

[0059] It should be noted that in order to adapt the anomaly metric to different scales of water areas and sensor distribution scenarios, the present invention will be implemented 、 、 Targeted calibration of parameters such as: Selection: It can be combined with the target water area’s characteristic frequency in the past period of time. The long-term monitoring statistics at the location are used to obtain the mean or median to ensure that it can represent the energy spectrum when it is relatively stable without intrusion or environmental interference. Here it is limited to arrive This is because in most inland waters or offshore scenes, waves and background noise usually fall at this level. Once the observed value is far above the upper limit or far below the lower limit, it often means an extreme situation and the database needs to be re-compared. Selection of: The range of 0.1 to 0.3 comes from the empirical value of trimodal fusion perception under normal circumstances. If the water area itself is relatively calm and there is not much noise interference, it can be selected around 0.1; if the environment is more complex or the monitoring point is relatively noisy, this template benchmark can be set at 0.2 to 0.3 to appropriately increase the sensitivity. Selection of the time constant: The main basis for the time constant of 2 to 10 seconds is the study of the diffusion rate of underwater acoustic, heat flow and electromagnetic disturbances at the meter level or even the hundred-meter level. If a certain water area has the characteristics of fast flow and the disturbance comes and goes quickly, it can be set to a smaller value to avoid delaying the response to short-term high-energy events; on the contrary, in a stable lake or reservoir, it is better to increase the time constant to a certain value. To prevent short-term small disturbances from triggering large-scale false alarms.

[0060] Through this formula, once If the water level continues to rise, the subsequent modules will consider that there are abnormal phenomena worthy of attention in the water environment; If the fluctuation is short-term or remains low, the adaptive alarm response unit will also maintain a low alarm intensity, which will not cause unnecessary resource consumption or erroneous intervention. It can serve as an important input for intrusion probability prediction and can also be archived for subsequent pattern recognition and statistical analysis during continuous monitoring. If the value is high and suspicious activity occurs within the corresponding water area, data around that time can be highlighted, helping operations and maintenance personnel investigate the source of the intrusion and optimize subsequent monitoring strategies. Because this formula integrates frequency domain energy spectrum deviation, fusion perception intensity differences, and temporal evolution, it is more flexible and accurate than traditional single threshold or simple fluctuation difference methods. Traditional solutions often only perform rough comparisons based on a single parameter or modality, which can easily fail due to multi-source noise or random water interference. This invention, however, integrates different physical quantities into a single measurement chain. Even if a particular modality is significantly affected by external factors, as long as other modalities or time decay characteristics can help distinguish authenticity, the system can more reliably return to accurate judgment. Furthermore, through careful design of the parameter range, the formula is both universal and adaptable in a variety of environments. This allows the intelligent water virtual electronic fence to effectively detect potential intrusions and provide reasonable anomaly measurements when deployed in freshwater rivers, coastal harbors, and even specific industrial water bodies.

[0061] Example 8: The invasion probability of each location in the target waters is predicted using the following formula:

[0062] ;

[0063] in, For location Location and current time The invasion probability ranges from 0 to 1; The center of the target water area; is the position uncertainty, dimensionless, ranging from 0.5 to 2; is the anomaly detection threshold, dimensionless, ranging from 1.5 to 3; is the initial detection probability; represents the Mahalanobis distance.

[0064] Specifically, It represents the system's prior probability of detection before any observation data or abnormal information is available. Its value is usually between 0.5 and 0.8. A larger value means that the system is more sensitive to potential threats, while a smaller value means that the system is relatively optimistic about the current environment. The risk assessment will only be significantly improved when obvious abnormal signs appear. Used to reflect the attenuation characteristics of spatial distribution, where Represents the spatial distance in the sense of Mahalanobis distance (or other equivalent distance metrics), which describes the monitoring points Relative to the center of the water area If the monitoring point is very close to the central area, the exponential term is close to 1, which means that the probability value here more directly reflects the local anomaly measurement and will be rapidly increased once a strong disturbance occurs. For points farther from the center, the exponential term will decay to a smaller value, thereby reducing the weight of distant points in the overall probability calculation. If the water area is large and the sensors are sparsely distributed, the attenuation of distance will be more prominent. If the water area monitoring center has multiple sub-centers, you can also define their own To achieve partition management. Third, Then the abnormality measure and anomaly detection threshold Ratio is performed, and this dimension reflects the consideration of the evolution of disturbance intensity in the present invention: when much higher than If If the value is still below or close to the threshold, it means that the abnormal signs are not enough to prove the existence of real intrusion. In actual deployment, the The range of 1.5 to 3 is selected to take into account the accuracy of distinguishing normal fluctuations and extreme disturbances. It reflects the ratio between the trimodal fusion perception strength and its template value, and is used to detect whether the energy level of the acoustic, optical and electrical signals exceeds the normal level at a certain moment and within a certain spatial distance. When the ratio is significantly higher than 1, it means that the potential disturbance has increased significantly, and the system has sufficient reason to suspect that the intruder is operating within this range. If the ratio is less than 1, it cannot be ruled out that it is caused by background noise or weak interference from the far end, and the risk is relatively low. Finally, The energy spectrum density of the wave described above is derived from the characteristic frequency Bandwidth-based template comparison: If the spectral density significantly exceeds the norm within a frequency range, it typically indicates an abnormal surge in surface fluctuations or internal vibrations, or high-energy disturbances caused by external interference. Once this ratio is superimposed, the water monitoring system can combine time-frequency characteristics with multimodal fusion observations for a more accurate comprehensive assessment.

[0065] In this way, the synergy of the above five product terms allows the entire invasion probability prediction process to not only have abnormal evolution information in the time dimension, but also integrate the attenuation characteristics of spatial coordinates and the overall characterization of water disturbances by multimodal perception signals, ultimately making the prediction results Constrained within the range of 0 to 1 for subsequent processing or alarm judgment. It is closer to the center, and at this moment 、 and When all of these factors are high, the product value will approach 1 due to the multiple amplification effect, and the system will determine that there is a high possibility of invasion at that point and that countermeasures must be initiated quickly; when any of these factors is significantly low, the product result will quickly fall back to a smaller value, indicating that the risk is not significant or is only a short-term disturbance, and the system can maintain a relatively low level of alert. It is worth noting that (Position uncertainty) values between 0.5 and 2 usually affect The calculation method of , especially when the water area sensor coverage is uneven or there is a certain error in the target position prediction in the flowing water area, the Mahalanobis distance can better handle the imbalance in different directions and scales; if If it is too large, it means that the system has a higher uncertainty about the target position, and it is necessary to moderately relax the threshold in the coordinate calculation. If it is too small, it means higher positioning accuracy, and spatial attenuation can more accurately present the diffusion of fluctuations within the water area.

[0066] Through such a set of calculation formulas that integrate multiple factors such as initial prior probability, spatial attenuation, anomaly measurement thresholding, three-modal fusion strength deviation, and spectral energy deviation, the present invention can provide an intuitive, comparable, and evolving intrusion probability value for each monitoring location at any time. It not only gets rid of the rigidity of traditional static threshold judgment, but also avoids the misjudgment or lack of sensitivity that may be caused by a single mode, so that water safety monitoring can still maintain high accuracy and real-time performance in large-scale and multi-interference scenarios. At the same time, such a probability distribution can also be used for hierarchical triggering of alarms: if the system finds that one or some points If the probability of intrusion is higher than a certain warning line, the monitoring and intervention efforts in that area will be increased accordingly; if the probability of intrusion in most locations is low over a period of time, the system can maintain a weak alarm output without completely shutting down, thus balancing resource efficiency and risk identification. 、 、 、 and Parameters such as these can be flexibly set according to the deployment environment and water characteristics, and can be dynamically calibrated through back-end algorithms or long-term historical data when necessary. Therefore, the formula has good scalability and adaptability, allowing the smart water virtual electronic fence to still work normally under different geographical conditions, seasonal changes, and even the coexistence of multiple interference sources. It can conduct reliable and detailed probabilistic assessments of various intrusion threats, and provide key risk quantification inputs for subsequent adaptive alarm response units.

[0067] Example 9: Calculate the current alarm intensity at each location using the following formula:

[0068] ;

[0069] in, For location Place and time The alarm intensity is dimensionless and ranges from 1 to 10; is the basic alarm intensity; is the critical intrusion probability.

[0070] Specifically, here The value of is limited to a dimensionless range of 1 to 10. The purpose is to provide the system with a convenient and externally explainable alarm level. The higher the value, the greater the alertness to intrusion at that location, which corresponds to more obvious sound, light or electromagnetic alarm means, and may even trigger additional security measures. This parameter is called the basic alarm intensity, which determines the lowest alarm level that the system can issue when it is in the initial state or low-risk state. The setting of is usually combined with the security needs of the target waters and the daily operating costs: when the environment is relatively sensitive or the risk of intrusion is high, the value can be relatively high to ensure that even small disturbances can be properly paid attention to; if the waters are relatively closed or the protection level is relatively loose, the value can be lowered to avoid unnecessary alarms and resource consumption. Ultimately, it's going to be on a scale of 1 to 10, so It is usually taken in the range of 1 to 5, which can not only maintain the basic warning function, but also leave enough room for adjustment for the gain or attenuation brought about by the subsequent multiplication factor.

[0071] Secondly, This partly reflects the nonlinear relationship between intrusion probability and alarm gain, where That is, the value given by the aforementioned intrusion probability prediction formula indicates that at position Place, time The possibility of an intrusion incident occurs under is the critical invasion probability. When the increase is only small and not enough to exceed the normal fluctuation level, the index item will The value is small and close to 0, so that the whole product approaches 0, which is reflected in the difference between the alarm intensity and the initial setting value. The difference is small; only when Begins to increase significantly and gradually approaches or exceeds hour, This design allows the system to not be overly alarmed when facing medium or low probability events, but to quickly "amplify" the alarm response in high probability situations, prompting managers to take necessary measures in a timely manner. The selection of the target waters is generally based on the historical intrusion risk data and the current security strategy. If the waters are relatively critical and cannot be lost, A slightly lower threshold (such as 0.7) makes the system more sensitive; if the environment can tolerate occasional non-lethal interference, the threshold can be increased (such as 0.9) to make the increase in alarm intensity more selective.

[0072] again, It is a time decay factor, which takes into account that the intensity of the alarm should be reduced when it lasts for a long time or gradually moves away from the time of the incident, so as to avoid the system maintaining a high intensity alarm after the threat has been eliminated or there is only a short fluctuation, thereby wasting resources or increasing management costs. Specifically, Usually set as the start time of this observation or abnormal event, (Characteristic development time constant) is used to reflect the "effective period" that the disturbance may actually last at the physical level. For example, in an underwater acoustic experiment, if an intrusion behavior can be determined as an abnormal disturbance or security threat on a time scale of about 2 to 10 seconds, then will also be taken in a similar range. Once and The time difference between Compared to the significant, then the attenuation factor will be significantly reduced, so that the alarm intensity gradually drops back to just slightly above The advantage of this is that once the system detects that the intrusion has gradually subsided or been dealt with, the alarm level will not remain high; but if the interference continues, It is not too large. At this time, the attenuation coefficient has not yet decayed deeply, and the alarm intensity can be maintained within a moderately rising range to ensure attention to persistent intrusions.

[0073] In summary, Responsible for "amplifying" the probability increment, and Focus on the time of "convergence" or "fallback"; both Multiplying these together, we ultimately create an adaptive alarm mechanism that can rise rapidly as the risk of intrusion increases, and gradually weaken over time. The value of is between 1 and 10. This invention ensures that the system remains controllable even in extreme situations: even if the probability of intrusion soars to an extremely high value, and the time decay is still in its early stages, the alarm intensity will not climb indefinitely, but will tend to saturate near the upper limit of 10. This can provide security managers with clear high-level signals without causing the system to fall into chaos due to excessive alarms. In practical applications, if certain locations If the probability of intrusion is high for a long time and has not been effectively checked, the alarm intensity will continue to approach 10 in a short period of time until the administrator intervenes and recalibrates the event time. Or intervene in the waters by other means; if the interference is only a momentary peak, then the invasion probability will drop rapidly, or After the interference has passed the main stage, the decay factor can allow the alarm intensity to naturally return to a lower level to avoid "long-lasting sounding" and disrupting normal operation. In addition, some deployment scenarios may require Raised relatively high and A smaller value is chosen so that any slightly suspicious interference can quickly lead to a significant alarm; in waters with relatively stable security conditions, a relatively loose initial condition can be configured so that the alarm intensity does not frequently jump to a high level in small interference events. It is worth noting that various external conditions such as hydrological seasonal changes, sensor coverage density, and environmental noise levels will affect the optimal configuration of the alarm strategy. Therefore, the present invention reserves the right to adjust the initial condition in the design. 、 as well as Dynamic adjustment space. When the system finds that false alarms or missed alarms occur frequently, it can try to adjust these key parameters in the next stage of operation through big data backtracking and parameter optimization. If the monitoring history shows that short-term spike anomalies are most likely to be related to real intrusions, the system can reduce Make the alarm decay more slowly; if the long-term high alarm is too frequent, Enlarge or moderately increase , so that only persistent disturbances with high confidence can maintain a high-level alarm. In general, through this alarm intensity formula, the present invention provides an efficient and flexible adaptive response method based on the perception and judgment of potential risks: on the one hand, it ensures that when a real threat occurs, the alarm can quickly increase to a value close to 10, so that on-site or remote managers are fully alert and respond in time; on the other hand, it can also quickly fall back when the risk is eliminated or there is only a slight disturbance, so as to control resource use and environmental interference within a reasonable range. The "probability judgment-alarm intensity-real-time attenuation" closed loop formed in this way provides a stable and scalable high-level decision-making capability for the entire set of intelligent water virtual electronic fences, and also allows the system to automatically adapt to the priorities of security events in a changing underwater environment, achieving the goals of reducing missed reports and false alarms while maintaining low energy consumption, low interference and high coverage.

[0074] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. An intelligent water area virtual electronic fence based on the fusion of sound, light and electricity, characterized by: It specifically includes: a water area sensor fusion network, a water area interface fluctuation feature extraction unit, an intrusion probability prediction unit, and an adaptive alarm response unit. The water area sensor fusion network is a multimodal perception network covering the target water area, consisting of acoustic sensors, optical sensors, and electromagnetic sensors, and is used to capture acoustic signals, optical signals, and electromagnetic signals in the target water area, and perform dynamic attenuation correction based on the physical properties of water to obtain a fusion perception result. The water area interface fluctuation feature extraction unit is used to extract the fluctuation energy spectrum density related to water body disturbance at various frequencies. The intrusion probability prediction unit is used to set a characteristic frequency, compare the fluctuation energy spectrum density at the characteristic frequency with the template fluctuation energy spectrum density at the characteristic frequency, and compare the fusion perception result with the template fusion perception result to calculate the current anomaly metric, set an initial detection probability, and predict the intrusion probability of each location in the target water area based on the anomaly metric, with the center of the target water area as the nominal point. The adaptive alarm response unit is used to set a basic alarm intensity and a critical intrusion probability, and calculate the current alarm intensity of each location based on the intrusion probability, and execute the early warning strategy based on the current alarm intensity of each location. The anomaly metric is calculated using the following formula: ； in, For time The abnormality measure of , dimensionless; is the characteristic frequency The wave energy spectral density at , in J / Hz; is the characteristic frequency The template fluctuation energy spectrum density at is in J / Hz and has a value range of arrive ; is the template fusion perception strength, dimensionless, ranging from 0.1 to 0.3; is the starting time, in seconds; is the characteristic development time constant, in seconds (s), with a value range of 2 to 10; For distance Place and time The fusion perception strength at this time is dimensionless and ranges from 0 to 1.

2. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 1 is characterized in that: The basic alarm intensity ranges from 1 to 5, dimensionless; the current alarm intensity ranges from 1 to 10, dimensionless; the critical intrusion probability ranges from 0.7 to 0.9, dimensionless; the initial detection probability ranges from 0.5 to 0.8, dimensionless.

3. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 2 is characterized in that: The early warning strategy includes: setting an early warning threshold, if the current alarm intensity of each location is greater than or equal to the early warning threshold, then issuing an early warning for the location, otherwise, then not issuing an early warning.

4. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 3 is characterized in that: The fusion perception result is calculated using the following formula: ； in, The initial intensity of the sound wave is in dB, ranging from 120 to 160; The initial intensity of the optical signal is in lux, ranging from 500 to 2000. is the initial strength of the electromagnetic signal, in V / m, ranging from 5 to 20; is the acoustic attenuation coefficient in water, which is related to water temperature and salinity, dimensionless, and ranges from 0.001 to 0.005; is the light signal attenuation coefficient in water, which is related to the water turbidity, dimensionless, and ranges from 0.05 to 0.2; is the electromagnetic wave attenuation coefficient in water, which is related to the water conductivity, dimensionless, and ranges from 0.01 to 0.1; The frequency of the sound wave is in Hz and ranges from 30 to 100; is the frequency of electromagnetic waves, in Hz, ranging from 1 to 10; is the refractive index of water, which is 1.33; is the speed of sound in water, in m / s, with a value range of 1450-1500; is the water density in kg / m³, with a value of 1000; The current time, in seconds.

5. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 4 is characterized in that: The wave energy spectral density is expressed using the following formula: ； in, Frequency The wave energy spectral density at , in J / Hz; is the effective height of the water surface wave, in m, with a value range of 0.01-0.1; is the acceleration due to gravity, in m / s², and its value is 9.8; is the peak frequency in Hz, ranging from 0.1 to 1; is the thermoacoustic coupling coefficient, dimensionless, ranging from 0.001 to 0.01; is the water temperature in K, ranging from 278 to 303; The conductivity of water is in S / m, ranging from 0.005 to 0.05, where S stands for Siemens; is the incident electric field strength, in V / m, ranging from 1 to 5; is the vacuum permeability, the unit is H / m, and the value is , H stands for Henry.

6. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 5, characterized in that: characteristic frequency The value range is 20-50, and the unit is Hz.

7. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 6, characterized in that: The invasion probability of each location in the target waters is predicted using the following formula: ； in, For location Location and current time The invasion probability ranges from 0 to 1; The center of the target water area; is the position uncertainty, dimensionless, ranging from 0.5 to 2; is the anomaly detection threshold, dimensionless, ranging from 1.5 to 3; is the initial detection probability; represents the Mahalanobis distance.

8. The intelligent water area virtual electronic fence based on the fusion of sound, light and electricity as claimed in claim 7 is characterized in that: The current alert intensity for each location is calculated using the following formula: ； in, For location Place and time The alarm intensity is dimensionless and ranges from 1 to 10; is the basic alarm intensity; is the critical intrusion probability.

Images