A water body environmental pollution detection method based on a marine high-precision salinity sensor

Abstract

This invention relates to the field of aquatic environmental pollution detection technology based on high-precision marine salinity sensors, and discloses a method for aquatic environmental pollution detection based on high-precision marine salinity sensors. It covers sensor network initialization, adaptive calculation of neighborhood radius, construction of a fixed spatial neighborhood set based on observation time series, extraction of local statistical features (mean, standard deviation, skewness, kurtosis), and construction of SMKI higher-order moment anomaly indices. Baseline analysis is performed using historical data, and anomaly judgment thresholds are adaptively set, ultimately achieving real-time anomaly detection and pollution event clustering identification of salinity data. Local neighborhood modeling improves detection sensitivity, the introduction of higher-order moment indices enhances the accuracy of anomaly identification, baseline modeling reduces false alarm and false negative rates, and density clustering supports structured output of pollution events.

Description

Technical Field

[0001] This invention relates to the field of water environment pollution detection technology based on high-precision marine salinity sensors, specifically to a water environment pollution detection method based on high-precision marine salinity sensors. Background Technology

[0002] With global climate change and overexploitation of marine resources, marine water pollution has become increasingly serious, posing a severe challenge to marine ecosystems and the sustainable development of human society. In recent years, marine environmental monitoring technology based on high-precision salinity sensors has received widespread attention, becoming an important means of detecting dissolved pollutants, assessing seawater structural stability, and assisting in early warning of marine ecological risks. However, existing pollution detection methods based on salinity sensors still have many shortcomings in practical applications, making it difficult to meet the needs for real-time, accurate, and sensitive detection in dynamic marine environments.

[0003] Traditional water pollution detection methods typically rely on empirical or fixed thresholds for alarms. A typical approach is to determine the upper and lower limits of salinity changes based on environmental experience or laboratory calibration; an alarm is triggered when the sensor detects salinity values ​​exceeding these limits. However, ocean salinity is influenced by multiple factors such as tides, waves, and temperature, resulting in complex spatiotemporal coupling characteristics. Using fixed thresholds easily leads to high false alarms and missed alarms. On the one hand, thresholds that are too low may misjudge normal seasonal or tidal fluctuations as pollution, increasing the burden on maintenance personnel; on the other hand, thresholds that are too high may fail to capture early or minor pollution events, delaying the location and control of pollution sources. To improve the adaptability of thresholds, some studies have used sliding window statistical methods or moving average filtering to dynamically update thresholds based on historical data. However, these methods only process first- or second-order statistics (such as mean and variance), failing to fully characterize the abrupt changes in distribution patterns caused by pollution events, and easily missing pollution signals with significant non-mean values ​​but sensitive distribution tail changes. Furthermore, the selection of the sliding window length and filtering window still requires empirical adjustment, lacking automated and theoretically supported optimization strategies. In multi-point distributed sensor networks, to balance spatial correlation, researchers employ spatial interpolation and kriging methods to reconstruct pollution features into a grid, or construct adjacency matrices based on graph theory for further graph convolution analysis. These methods can utilize spatial neighborhood information to enhance the spatial identifiability of anomalies. However, in the complex topological structure of the ocean at large scales, interpolation computations are computationally intensive and sensitive to grid resolution. Anomaly detection models based on graph convolution or deep learning, while highly accurate, require a large number of labeled samples and expensive computational resources, making them unsuitable for resource-constrained ocean buoys or automated observation stations. Some scholars have also introduced multivariate indicators such as information entropy, fractal dimension, and multi-scale analysis to identify salinity anomalies. However, most methods focus on overall distribution feature changes, lacking precise characterization of local fine-grained regional distribution patterns, making it difficult to achieve early warning of single-point or small-scale pollution events. Furthermore, these techniques often use features of different dimensions as input, then fuse thresholds through rules or models for judgment, resulting in a lengthy process with numerous parameters, which is not conducive to real-time online deployment.

[0004] Therefore, this paper aims to propose a water pollution detection method based on a high-precision marine salinity sensor. It captures local distribution features by adaptively constructing spatial neighborhoods, uses third-order and fourth-order central moments to construct skewness and kurtosis to measure high-order abnormal behavior, and then uses SMKI to identify anomalies. Finally, neighborhood density clustering is used to reconstruct the complete evolution process of pollution events. Summary of the Invention

[0005] This invention provides a method for detecting water pollution based on a high-precision marine salinity sensor, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: a method for detecting water pollution based on a high-precision marine salinity sensor, comprising:

[0007] Initialize the sensor network and adaptively calculate the spatial neighborhood radius of each sensor;

[0008] Within a preset observation time series, the salinity observation values ​​of each sensor at each time point are obtained, and a corresponding fixed spatial neighborhood set is constructed based on the spatial neighborhood radius.

[0009] Based on the spatial neighborhood set, calculate the mean and standard deviation of salinity within the local region;

[0010] The third and fourth central moments within the local region are further calculated, and the skewness and kurtosis values ​​are obtained accordingly.

[0011] The local higher-order moment index SMKI is constructed based on skewness and kurtosis indices;

[0012] Baseline analysis was performed based on historical observation data, and an adaptive method was used to calculate the anomaly detection threshold for each sensor.

[0013] Anomaly detection and identification of outliers are performed on real-time observation data using constructed SMKI indicators and thresholds.

[0014] Anomaly clusters are identified using a density clustering method based on spatial neighborhood, and pollution event information is output based on the spatial-temporal correlation of the anomaly clusters.

[0015] Optionally, the initialization of the sensor network and adaptive calculation of the spatial neighborhood radius of each sensor specifically includes:

[0016] The distance between the i-th sensor and the p-th sensor is denoted as d. i,p ;

[0017] Where i,p∈{1,2,...,N} are sensor numbers; N is the total number of sensors;

[0018] Construct an N×N dimensional real symmetric distance matrix D = [d i,p ] N×N ; where d i,p =d p,i d i,i =0;

[0019] Calculate the minimum distance between sensor i and all other sensors.

[0020] Calculate the spatial neighborhood radius The median(·) function returns the median value of an ordered list.

[0021] Optionally, the step of acquiring the salinity observation values ​​of each sensor at each time point within a preset observation time series, and constructing a corresponding fixed spatial neighborhood set based on the spatial neighborhood radius, specifically includes:

[0022] Set the observation time set as {t} j |j=1,2,...,M};where t j Let j be the time point of the j-th observation; M is the total number of observation times; and j is the number of the observation time point.

[0023] Obtain the value of sensor i at time t j The salinity value is denoted as S. i,j ;

[0024] Construct the spatial neighborhood number set of sensor i.

[0025] Optionally, the step of calculating the mean and standard deviation of salinity within a local region based on a spatial neighborhood set specifically includes:

[0026] For any i∈{1,2,...,N}, j∈{1,2,...,M}, perform the following steps:

[0027] S101. Calculate the neighborhood size Where |·| is the set cardinality operation;

[0028] S102. Calculate the neighborhood of the i-th sensor at time t. j arithmetic mean of salinity

[0029] S103. Calculate the neighborhood of the i-th sensor at time t. j Salinity standard deviation σ i,j :

[0030]

[0031] Optionally, the further calculation of the third and fourth central moments within the local region, and the determination of skewness and kurtosis values ​​accordingly, specifically includes:

[0032] Calculate the third-order center distance

[0033] Calculate skewness

[0034] Calculate the fourth-order center distance

[0035] Calculate kurtosis

[0036] Optionally, the construction of the local higher-order moment index SMKI based on skewness and kurtosis indices specifically includes:

[0037] Constructing the local higher-order moment index SMKI i,j =|γ i,j |+|κ i,j |; where |·| is the absolute value.

[0038] Optionally, the step of performing baseline analysis based on historical observation data and calculating the anomaly detection threshold for each sensor using an adaptive method specifically includes:

[0039] Select the first B time points {t1,t2,...,t B} as the baseline period, and 1≤B <M;

[0040] Set the anomaly detection threshold for the i-th sensor to be...

[0041] Optionally, the step of using the constructed SMKI index and threshold to perform anomaly detection on real-time observation data and identify anomalies specifically includes:

[0042] For any j>B: if SMKI i,j >H i Then mark the i-th sensor at time t. j This is an outlier.

[0043] Optionally, the step of identifying anomalous clusters using a density clustering method based on spatial neighborhood and outputting pollution event information based on the spatial-temporal correlation of the anomalous clusters specifically includes:

[0044] Extraction time t j The abnormal sensors are numbered and grouped into set A. j ={i|SMKI i,j >H i};

[0045] Perform the following steps to perform point-by-point iterative neighborhood clustering:

[0046] S2100, i∈A j All are marked as ungrouped;

[0047] S2200. Take an ungrouped point i in sequence, create a new contamination cluster C = {i}, and mark i as a grouped point;

[0048] S2300. For each cluster member p∈C, perform the following steps:

[0049] S2310, Traverse the remaining ungrouped points q∈A j :

[0050] S2311, if dp,q If ≤r, then add q to C and mark it as a group;

[0051] S2400, Repeat step S2300 until there are no new members in cluster C; Record C as a final cluster;

[0052] S2500, Repeat steps S2100 to S2400 until all points are grouped;

[0053] If time t j-1 Cluster C j-1 With t j Cluster C j There is an intersection of the numbers, that is This is considered as a continuation of an event, and its time intervals are merged;

[0054] For each final event cluster C, output:

[0055] Sensor list {i|i∈C}; time tag interval between first and last occurrences [t] min ,t max Peak index

[0056] The present invention has the following beneficial effects:

[0057] 1. By constructing a real symmetric distance matrix, the spatial relationships between sensors are quantified, and a strategy is proposed to adaptively determine the neighborhood radius using the "minimum median distance." First, this method does not rely on prior knowledge or manually set the neighborhood radius, solving the detection performance degradation problem caused by unreasonable neighborhood settings in traditional methods and enhancing the system's adaptability to sensor networks with different distribution densities. Second, by using the minimum median distance as the estimation method for the global scale, the interference of abnormal nodes or dense clusters on the radius setting is effectively reduced, exhibiting good robustness. Compared with existing methods based on fixed radii or empirical values, this method automatically adjusts the neighborhood size, maintaining high regional coverage and detection accuracy even in cases of uneven spatial distribution or node damage. Furthermore, this neighborhood partitioning mechanism lays a solid foundation for subsequent local statistical analysis and spatial clustering.

[0058] 2. By uniformly constructing the set of observation times and the set of spatial neighborhoods, a stable data input structure is formed, providing clear boundaries for subsequent algorithm design. Synchronizing the construction of temporal information and spatial neighborhoods ensures that the entire data structure possesses "spatiotemporal synchronization." This mechanism differs from existing methods that only focus on time series or spatial discrete points, providing higher resolution for processing spatiotemporally coupled data. Furthermore, the construction of spatial neighborhoods is based on the adaptive radius determined in step 1, ensuring that the neighborhood of each sensor accurately reflects its local water characteristics, thus avoiding local distortion caused by "too wide" or "too narrow" neighborhoods. Compared to traditional global processing strategies such as sliding windows, this method employs a local observation mechanism, making it more suitable for handling localized anomalous pollution events, especially in scenarios with dispersed pollution sources in large-scale water bodies.

[0059] 3. Based on data within a fixed spatial neighborhood, the mean and standard deviation of local salinity are calculated to form basic statistical characteristics. This "spatial localization" statistical construction enables each node to possess personalized background modeling capabilities at the time of observation, avoiding generalization errors caused by using global thresholds. Traditional detection methods based on global mean or sliding window standard deviation often overlook local characteristics, easily misidentifying natural fluctuations as anomalies or missing weak anomalies. This method reflects the range of local fluctuations through neighborhood standard deviation, providing a standardized benchmark for higher-order anomaly indicators (skewness, kurtosis), ensuring consistent scale in indicator calculations. This improves the sensitivity of detecting small-scale pollution mutations and adapts to the diverse changes in complex marine dynamic environments.

[0060] 4. Introducing skewness and kurtosis analysis to analyze the local data distribution pattern is one of the core technologies of the entire method. Compared with traditional strategies that rely solely on mean / variance to identify anomalies, skewness characterizes the asymmetry of the distribution, and kurtosis reflects extreme values ​​or peak anomalies, enabling the capture of weak, non-mean-based abrupt changes. Higher-order moments from statistics are introduced into sensor network data analysis to improve the ability to identify anomalies of the "distribution variation" type. Simultaneously, skewness and kurtosis are dimensionless, supporting comparative analysis across time and nodes, enhancing the model's versatility. Compared with existing methods, the use of these higher-order moments not only improves detection accuracy but also greatly expands the types of detectable anomalies (such as mild anomalies and tail contamination), making it invaluable in early pollution warning.

[0061] 5. SMKI is a comprehensive anomaly index proposed in this scheme, which combines the absolute values ​​of skewness and kurtosis to form a single anomaly measure. It simplifies two high-order features into a unified standard, simplifying subsequent judgment logic and improving processing efficiency; at the same time, it avoids complex multi-dimensional decisions in high-dimensional feature spaces, improving real-time performance. Theoretically, SMKI integrates asymmetry and peaked features, providing a comprehensive characterization of local "distribution extremity," taking into account both amplitude and morphology. Compared to traditional methods such as z-score, which rely only on first- or second-order statistics, SMKI has stronger expressive power. In practical applications, it can accurately identify various types of water anomalies such as "pollution peaks," "offset diffusion," and "local high-frequency anomalies," providing possibilities for pollution event classification and identification.

[0062] 6. By selecting historical observation period data (baseline period) to set anomaly thresholds, the system is entirely based on historical normal state data and does not introduce any prior model parameters, representing a typical "data-driven adaptive learning" strategy. The baseline period length can be flexibly selected, supporting periodic modeling of different dynamic marine environments (such as diurnal tidal cycles); simultaneously, the use of a maximum value strategy to construct the thresholds naturally provides high safety redundancy. Compared to common fixed threshold and empirical coefficient setting methods, the adaptive mechanism, while ensuring false alarm rate control, is better able to cope with systemic changes brought about by dynamic backgrounds (such as changes in temperature difference, wind speed, and water flow). Furthermore, the thresholds are entirely data-driven, possessing transparency and interpretability, making them suitable for deployment in unattended marine intelligent monitoring platforms.

[0063] 7. Combining SMKI and adaptive thresholding, this step enables real-time anomaly identification, a crucial decision-making step in the algorithm execution. Employing a "point-by-point judgment + local thresholding" method overcomes the real-time and accuracy bottlenecks of traditional "overall window statistics." Anomaly detection does not require waiting for a time window to complete, allowing for immediate response to sudden events, significantly reducing pollution response delays. This is particularly suitable for tasks such as sudden leak monitoring and early red tide identification.

[0064] 8. Combining outliers into "contamination clusters" using density clustering based on spatial neighborhood, and then constructing a complete pollution event through temporal merging, is a crucial transition from "point anomalies" to "structural events." This method employs unsupervised clustering based on neighborhood propagation, independent of external parameters such as cluster size or shape assumptions, and fully self-organizes the identification of pollution patterns. In real-world scenarios, pollution sources often exhibit simultaneous changes at multiple points and spread over time. Traditional methods relying solely on threshold alarms cannot track the full picture of an event. This method, however, achieves dynamic reconstruction of the pollution "start-spread-terminus" process through synchronous tracking in both spatiotemporal domains, providing comprehensive data support for subsequent decisions such as source tracing, alarm response, and containment. Attached Figure Description

[0065] Figure 1This is a schematic diagram of the process of the present invention. Detailed Implementation

[0066] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0067] Example, refer to Figure 1 A method for detecting water pollution based on a high-precision marine salinity sensor, comprising:

[0068] Initialize the sensor network and adaptively calculate the spatial neighborhood radius of each sensor;

[0069] Within a preset observation time series, the salinity observation values ​​of each sensor at each time point are obtained, and a corresponding fixed spatial neighborhood set is constructed based on the spatial neighborhood radius.

[0070] Based on the spatial neighborhood set, calculate the mean and standard deviation of salinity within the local region;

[0071] The third and fourth central moments within the local region are further calculated, and the skewness and kurtosis values ​​are obtained accordingly.

[0072] The local higher-order moment index SMKI is constructed based on skewness and kurtosis indices;

[0073] Baseline analysis was performed based on historical observation data, and an adaptive method was used to calculate the anomaly detection threshold for each sensor.

[0074] Anomaly detection and identification of outliers are performed on real-time observation data using constructed SMKI indicators and thresholds.

[0075] Anomaly clusters are identified using a density clustering method based on spatial neighborhood, and pollution event information is output based on the spatial-temporal correlation of the anomaly clusters.

[0076] From sensor initialization and adaptive neighborhood construction to local statistical feature extraction, anomaly index calculation, adaptive threshold setting, real-time anomaly identification, and finally to the clustering output of pollution events, a closed-loop detection system is formed. By comprehensively integrating information from both spatial and temporal dimensions, it solves the problems of highly dynamic data, hidden local pollution events, and the inapplicability of global models in marine environments. Local neighborhood calculation is used to improve sensitivity, higher-order moment indices are introduced to enhance anomaly identification accuracy, baseline modeling avoids false alarms and false negatives, and finally, clustering identification achieves quantitative output of pollution events. Compared to existing static threshold alarms or sliding window methods, this solution balances detection accuracy, response timeliness, and structural interpretability, making it particularly suitable for water quality anomaly monitoring tasks in the variable and dynamic marine environment.

[0077] The initialization of the sensor network and adaptive calculation of the spatial neighborhood radius of each sensor specifically includes:

[0078] The distance between the i-th sensor and the p-th sensor is denoted as d. i,p ;

[0079] Where i,p∈{1,2,...,N} are sensor numbers; N is the total number of sensors;

[0080] Construct an N×N dimensional real symmetric distance matrix D = [d i,p ] N×N ; where d i,p =d p,i d i,i =0;

[0081] It provides pairwise distance relationships between all sensors, laying the foundation for subsequent neighborhood division and spatial statistics;

[0082] Calculate the minimum distance between sensor i and all other sensors. The minimum spacing between each sensor is evaluated to characterize the typical spatial distribution between measurement points;

[0083] Calculate the spatial neighborhood radius The median(·) function returns the median value of an ordered list. The neighborhood radius is automatically generated based on the median minimum distance of all sensors, avoiding manual setting and improving the robustness of the algorithm.

[0084] By calculating the minimum distance between sensors and using the median of the minimum distances between all sensors to set the spatial neighborhood radius, adaptive setting of neighborhood parameters is achieved. This approach solves the problems of unsuitability and local distortion caused by relying on empirical parameters or globally fixed values ​​in traditional neighborhood partitioning methods. The advantages of this method lie in its "adaptability" and "robustness": using the median in the distance matrix as a representative scale not only automatically reflects the spatial density changes of sensors in actual deployment but also copes with situations such as partial node failures or irregular distribution, ensuring that the neighborhood construction is realistic and stable. This design effectively improves the representativeness and accuracy of subsequent statistical indicators, laying a stable spatial structural foundation for the entire pollution detection system and is a key prerequisite for achieving localized detection and distribution-sensitive detection capabilities.

[0085] Within a preset observation time series, the salinity observation values ​​of each sensor at each time point are acquired, and a corresponding fixed spatial neighborhood set is constructed based on the spatial neighborhood radius. Specifically, this includes:

[0086] Set the observation time set as {t} j |j=1,2,...,M};where t j Let j be the time point of the j-th observation; M is the total number of observation times; and j is the number of the observation time point.

[0087] Obtain the value of sensor i at time t j The salinity value is denoted as S. i,j ;

[0088] Define the time required by the algorithm and the input format of the measured values ​​to provide a data source for subsequent statistical calculations;

[0089] Construct the spatial neighborhood number set of sensor i. Determine the local neighborhood range of each sensor to provide a set of objects for subsequent local statistics.

[0090] The organization of observation time was clearly defined, and a fixed spatial neighborhood set was constructed for each sensor based on the previously calculated spatial neighborhood radius. This structure organizes the data as a combination of "spatiotemporal points + neighborhood structure," providing an efficient data access framework for subsequent calculations. By constructing a fixed neighborhood set, the problem of ineffective statistical analysis caused by the spatial discreteness and inconsistent dimensionality of the original observation data is solved. The advantage of this operation is that each moment and each sensor has contextual neighborhood support, thereby realizing true local statistical analysis. This approach provides the necessary data support for higher-order statistics such as local mean, standard deviation, skewness, and kurtosis, ensuring the feasibility of the entire higher-order moment anomaly analysis method. At the same time, the introduction of fixed neighborhoods also greatly improves data processing efficiency, making the algorithm suitable for deployment on resource-constrained edge computing nodes or embedded devices.

[0091] The calculation of the mean and standard deviation of salinity within a local region based on a spatial neighborhood set specifically includes:

[0092] For any i∈{1,2,...,N}, j∈{1,2,...,M}, perform the following steps:

[0093] S101. Calculate the neighborhood size Where |·| is the set cardinality operation; it calculates the number of sensors in the neighborhood, providing a normalization factor for subsequent average and variance calculations;

[0094] S102. Calculate the neighborhood of the i-th sensor at time t. j arithmetic mean of salinity The neighborhood of the i-th sensor at time t is obtained. j The average salinity level describes the local background concentration;

[0095] S103. Calculate the neighborhood of the i-th sensor at time t. j Salinity standard deviation σ i,j :

[0096] The fluctuation range of neighborhood salinity is quantified to provide a standardized benchmark for skewness and kurtosis calculation.

[0097] By calculating the mean and standard deviation of salinity observations within the neighborhood of each sensor, a basic statistical framework for assessing local environmental conditions is constructed. The "mean" characterizes the local background level, and the "standard deviation" delineates the degree of local fluctuation, providing a standardized quantitative basis for subsequent higher-order analysis and anomaly detection. This effectively solves the problem of traditional global models failing due to drastic fluctuations in sensor readings caused by local hydrodynamics, temperature gradients, and other factors in marine environments. Compared to global mean or moving average methods, this local mean / standard deviation system exhibits stronger sensitivity, particularly in detecting regional pollutant leaks and localized red tide anomalies. Furthermore, the introduction of standard deviation allows subsequent skewness and kurtosis to be normalized to a dimensionless form, improving the comparability of indicators and serving as a fundamental step in achieving higher-order anomaly detection.

[0098] The further calculation of the third and fourth central moments within the local region, and the subsequent determination of skewness and kurtosis values, specifically includes:

[0099] Calculate the third-order center distance Measure the skewness of the neighborhood salinity distribution; extract information on the skewness of the local salinity distribution to initially capture the asymmetry of the distribution;

[0100] Calculate skewness The third central moment is standardized into a dimensionless skewness, which facilitates comparison between different regions or time periods.

[0101] Calculate the fourth-order center distance Measure the kurtosis or thick tail of the salinity distribution in the neighborhood; extract the kurtosis information of the local salinity distribution and measure the thickness of the tail or the degree of peaking;

[0102] Calculate kurtosis After subtracting 3, it aligns with the normal distribution; the fourth central moment is transformed into excess kurtosis, which is horizontally aligned with the kurtosis of the normal distribution, highlighting the abnormal tail characteristics.

[0103] This paper details how to calculate third- and fourth-order central moments based on neighborhood data, and derive skewness and kurtosis from these as core indicators for measuring the distribution pattern of local data. Through this step, the system can identify distribution asymmetry (skewness) and anomalous peaks (kurtosis), enabling the identification of subtle anomalies or extreme pollution behaviors that traditional mean / variance methods struggle to capture. This solves the problem of "Gaussian assumption failure" caused by traditional anomaly detection algorithms relying solely on first- or second-order statistics, significantly improving the sensitivity of detection to asymmetric or heavy-tailed pollution events. For example, in marine micro-leaks or sudden spread scenarios, the pollution signal may not significantly increase the average value, but its distribution will exhibit a clear skewness or concentration trend; in this case, skewness / kurtosis is crucial for identification. By combining these two higher-order statistical indicators, not only are the detection dimensions enhanced, but a solid foundation is also laid for subsequent indicator fusion and cluster analysis.

[0104] Constructing a local high-order moment index SMKI based on skewness and kurtosis indexes specifically includes: :构造局部高阶矩指标SMKI,具体包括:

[0105] Constructing the local high-order moment index SMKI :构造局部高阶矩指标SMKI i,j =|γ i,j |+|κ i,j |; where |·| is the absolute value; :=|γ|+|κ|;其中,|·|为绝对值;

[0106] Combining the absolute values of skewness and kurtosis to form a single metric that directly reflects the degree of abnormality in the neighborhood high-order distribution. :联合偏度与峰度绝对值,形成单一度量,直接反映邻域高阶分布异常程度。

[0107] A method for constructing the SMKI index is proposed. By adding the absolute values of skewness and kurtosis, a unified high-order anomaly detection index is formed. This design overcomes the problem of limited coverage of a single index by integrating information on two distribution characteristics, achieving a wider anomaly detection ability. It improves the real-time performance and computational efficiency of the detection algorithm. Compared with multi-dimensional vector discrimination, this single index can be used for rapid threshold judgment and is suitable for embedded environments and online real-time detection scenarios; it improves the accuracy of anomaly recognition. Skewness is sensitive to distribution skewness and kurtosis is sensitive to anomaly concentration. By means of a unified index, the collaborative anomalies of both can be captured. :提出了SMKI指标的构造方法,通过将偏度和峰度的绝对值相加,形成一个统一的高阶异常检测指标。该设计通过整合两种分布特性信息,克服了单一指标覆盖范围有限的问题,实现了更广泛的异常检测能力。提升检测算法的实时性与计算效率。相比于多维向量判别,该单一指标可用于快速阈值判断,适合嵌入式环境与在线实时检测场景;提升异常识别的精度。偏度敏感于分布倾斜、峰度敏感于异常集中,通过统一指标可捕捉两者协同异常。

[0108] Performing a baseline period analysis based on historical observation data and calculating the anomaly determination threshold for each sensor using an adaptive method specifically includes: :所述根据历史观测数据进行基线期分析,采用自适应方法计算各传感器的异常判定阈值,具体包括:

[0109] Selecting the first B moments {t1, t2,..., t B} as the baseline period, and 1 ≤ B < M; defining a pollution-free reference period for generating an adaptive threshold; if the system or environment has an obvious intraday cycle (such as tides, diurnal temperature differences, etc.), the baseline period should cover at least one complete cycle. For example, if the salinity sensor samples once an hour and the diurnal cycle is 24 hours, then B ≥ 24 can be taken. The larger B is, the more historical maxima are included, the higher the threshold, and the lower the sensitivity of the system to minor anomalies; conversely, the smaller B is, the lower the threshold, the easier it is to trigger an alarm, but the more likely false alarms are to occur. :选取前B个时刻{t1,t2,...,t}作为基线期,并且1≤B<M;划定无污染的参考时段,用于生成自适应阈值;若系统或环境具有明显 的日内周期(如潮汐、昼夜温差等),应让基线期至少覆盖一个完整周期。例如盐度传感器每小时采样一次,昼夜周期为24小时,则可取B≥24。B越大,包含的历史极大值越多,阈值越高,系统对轻微异常的敏感性越低;反之,B越小,阈值越低,更易触发报警,但也更容易出现误报。

[0110] Setting the anomaly determination threshold for the i-th sensor as :设置第i号传感器的异常判定阈值为 Directly taking the maximum value of the index within the baseline period as the threshold without any external parameters, ensuring that the threshold is completely determined by historical normal data. :直接取基线期内指标最大值作为阈值,无任何外部参数,保证阈值完全由历史正常数据决定。

[0111] By introducing a baseline period mechanism, a dynamic, parameter-free anomaly detection threshold is constructed using historical data, thus avoiding the subjectivity and low adaptability issues caused by manual settings. The maximum SMKI value observed during the baseline period is directly used for the current threshold setting, ensuring the accuracy and stability of detection. This solves the problem of false alarms and missed alarms caused by dynamic environmental changes (such as tides, temperature, and sunshine) to fixed threshold detection systems, significantly improving the adaptive capability and practical application range of the detection algorithm. The system has a "learning" capability, automatically adjusting its sensitivity based on the historical state of each sensor to achieve a "location-specific" monitoring strategy, which is a key component for achieving highly reliable online monitoring and alarm.

[0112] The method of using the constructed SMKI index and threshold to perform anomaly detection on real-time observation data and identify anomalies specifically includes:

[0113] For any j>B: if SMKI i,j >H i Then mark the i-th sensor at time t. j It identifies abnormal points; it determines online whether the current distribution pattern exceeds the historical normal range, and realizes real-time pollution alarm.

[0114] This paper proposes a specific discrimination logic for identifying anomalies based on the relationship between SMKI (Smart Water Quality Indicator) and thresholds within a real-time monitoring framework. Each moment and each node can independently determine the presence of anomalies without waiting for window data or post-processing, significantly improving detection response speed. This real-time mechanism solves typical problems in traditional systems such as "lagging response" and "discontinuous alarms," ​​especially in the early stages of pollution diffusion, enabling immediate anomaly identification, triggering early warning and response mechanisms, and preventing pollution from escalating. Furthermore, this method integrates complex indicators with simple decision-making, providing feasibility for deployment on resource-constrained platforms such as marine buoys and autonomous surface vessels, thus contributing to the practical implementation of intelligent water environment monitoring technology.

[0115] The method of identifying anomalous clusters using a spatial neighborhood-based density clustering approach and outputting pollution event information based on the spatial-temporal correlation of these anomalous clusters specifically includes:

[0116] Extraction time t j The abnormal sensors are numbered and grouped into set A. j ={i|SMKI i,j >H i}; Summarize all abnormal sensor numbers at the current moment to prepare for clustering identification;

[0117] Perform the following steps to perform point-by-point iterative neighborhood clustering:

[0118] S2100, i∈A j All are marked as ungrouped;

[0119] S2200. Take an ungrouped point i in sequence, create a new contamination cluster C = {i}, and mark i as a grouped point;

[0120] S2300. For each cluster member p∈C, perform the following steps:

[0121] S2310, Traverse the remaining ungrouped points q∈A j :

[0122] S2311, if d p,q If ≤r, then add q to C and mark it as a group;

[0123] S2400, Repeat step S2300 until there are no new members in cluster C; Record C as a final cluster;

[0124] S2500, Repeat steps S2100 to S2400 until all points are grouped;

[0125] Based on spatial neighborhood relationships, anomalies are divided into pollution clusters, and the spatial range of each cluster is defined.

[0126] If time t j-1 Cluster C j-1 With t j Cluster C j There is an intersection of the numbers, that is This means treating it as a continuation of an event and merging its time intervals; tracking the evolution of the same pollution cluster across time periods to achieve a closed-loop event time sequence;

[0127] For each final event cluster C, output:

[0128] Sensor list {i|i∈C}; time tag interval between first and last occurrences [t] min ,t max Peak index Generate complete pollution incident reports to provide accurate information for decision-making and response.

[0129] This method utilizes density clustering algorithms to organize outliers in a spatial neighborhood into pollution clusters, and combines this with temporal relationships to identify the evolution of pollution events. Employing an unsupervised, neighborhood-propagation-based iterative algorithm effectively avoids the reliance of traditional clustering on assumptions about cluster numbers or structure. It solves the core problem of scattered outlier distribution and difficulty in determining whether an event constitutes a pollution event, significantly improving pollution identification from "point identification" to "event-level understanding." It can output specific pollution sensor data, duration, and intensity indicators, providing regulatory agencies with actionable reports and decision-making support, and also providing structured foundational data for pollution source tracing and dynamic response. This serves as a bridge between the technical and operational systems of this solution.

[0130] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0131] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A water body environmental pollution detection method based on a marine high-precision salinity sensor, characterized in that, The method comprises the following steps: initializing a sensor network and adaptively calculating a spatial neighborhood radius of each sensor, comprising: Get the Sensor No. and No. The distance between the sensors is denoted as . ; wherein, is the sensor number; is the total number of sensors; Constructing Virtually symmetric distance matrix ; wherein, , ; The minimum distance between the first sensor and all other sensors is calculated ;​ Computing spatial neighborhood radius ; wherein, is the median function that returns the middle value of an ordered list; acquiring a salinity observation value of each sensor at each time point within a preset observation time sequence, and constructing a corresponding fixed spatial neighborhood set according to the spatial neighborhood radius, comprising: The set of observation time instants is set ; wherein is the time point of the th observation; is the total number of observation time instants; is the number of the observation time point; acquire the salinity value of the first numbered sensor at the time instant , denoted as ; Constructing a first number set of spatial neighborhood numbers of the sensor ; calculating a mean value and a standard deviation of salinity in a local region based on the spatial neighborhood set, comprising: for any , , the following steps are performed: S101、Calculate neighborhood size ; wherein, is a set cardinality operation; S102、calculating the first arithmetic mean of the salinity of the sensor neighborhood at the time point in time ; S103、calculating the standard deviation of the salinity of the neighborhood of the sensor at the time :​​ ； further calculating a third-order central moment and a fourth-order central moment in the local region, and obtaining skewness and kurtosis values respectively based on the third-order central moment and the fourth-order central moment, comprising: Computing third order center distance ; Computing skewness ; Computing fourth order center distance ; Computing kurtosis ; constructing a local high-order moment index SMKI according to the skewness and kurtosis indexes, comprising: Constructing a local higher moment indicator ; wherein is an absolute value; performing baseline period analysis according to historical observation data, and adaptively calculating an abnormality judgment threshold of each sensor, comprising: Selecting the previous moment in time as a baseline period, and ; Set the first The anomaly detection threshold for sensor number is ; performing abnormality detection on real-time observation data by using the constructed SMKI index and the threshold, and identifying abnormal points; identifying an abnormal cluster by a density clustering method based on a spatial neighborhood, and outputting pollution event information according to a space-time correlation of the abnormal cluster.

2. The water environmental pollution detection method based on the marine high-precision salinity sensor according to claim 1, characterized in that, The abnormality detection on real-time observation data by using the constructed SMKI index and the threshold, and the identification of abnormal points, specifically comprise: For any : If , mark the th sensor as an outlier at time . 3.The water environmental pollution detection method based on the marine high-precision salinity sensor according to claim 2, characterized in that, The identification of an abnormal cluster by a density clustering method based on a spatial neighborhood, and the outputting of pollution event information according to a space-time correlation of the abnormal cluster, specifically comprise: extracting the time instant of the abnormal sensor number and composing a set ; performing the following steps to perform point-by-point iterative neighborhood clustering: S2100、to All are marked as ungrouped; S2200, take one ungrouped point , create a new pollution cluster , and mark as grouped; S2300, for each cluster member performing the following steps: S2310, traverse the remaining ungrouped points : S2311、If then add to mark it as grouped; S2400, repeat step S2300 until there are no new members in the cluster ; record as a final cluster; S2500, repeating steps S2100 to S2400 until all points are grouped. If the time cluster with cluster exists a number intersection, i.e. , it is considered as an event continuation, merging its time interval. for each final event cluster , output: Sensor list ; first and last occurrence time tag interval ; peak indicator .

Images