A sample target data enhancement processing method for transient electromagnetic surface detection

Abstract

The present invention provides a sample target data enhancement processing method for transient electromagnetic surface detection, which relates to the technical field of sample target data enhancement processing. The method comprises the following steps: uniformly dividing a detection area into feature unit grids, using the center point of each grid as a monitoring point, and collecting transient electromagnetic signal data; judging whether a monitoring point signal is missing, and if missing, defining it as a data missing point and filling it based on a spatial distance attenuation law; using a multi-dimensional anomaly detection algorithm to identify noise interference monitoring points, setting a neighborhood search radius, and correcting it using the signal value of a non-noise monitoring point; constructing a BIM model of the target detection area, performing registration and real-time updating of monitoring point data, defining a dynamic risk threshold, and triggering an unexploded ordnance warning signal if the monitoring point signal exceeds the threshold. The present invention achieves comprehensive monitoring and risk assessment of the monitoring area.

Description

Technical Field

[0001] The present invention relates to the technical field of sample target data enhancement processing, and in particular to a sample target data enhancement processing method for transient electromagnetic surface detection. Background Art

[0002] In the detection of unexploded ordnance (UXO), transient electromagnetic surface detection technology is widely used to identify dangerous objects buried in the soil. However, current technology has significant deficiencies in data acquisition and processing, particularly regarding missing signal data at monitoring points. Data omissions during the monitoring process can lead to missed detection of UXO, posing a serious safety hazard in practical applications. Data omissions are typically caused by a variety of factors, including environmental noise, electromagnetic interference, and equipment failure, which compromise signal integrity. In UXO detection, any missing data can lead to an incorrect assessment of potential threats, delaying bomb disposal efforts and increasing the risk of accidents.

[0003] Furthermore, existing technologies also have limitations in identifying and correcting for noise interference. Existing methods typically rely on fixed thresholds or simple filtering techniques, which make it difficult to effectively distinguish normal signals from noise interference in complex geographical and environmental conditions. As a result, some real signals may be mistakenly labeled as noise and ignored, increasing the probability of missing unexploded ordnance. Furthermore, existing data processing technologies lack real-time performance and cannot quickly respond to changes in monitoring data, resulting in less timely and effective assessments of potential risks. This lag not only affects detection accuracy but also limits the ability to quickly respond to dynamic risks. Therefore, to meet the needs of unexploded ordnance detection, there is an urgent need to improve existing technologies to achieve more efficient and accurate data collection and processing, ensuring that no potential hazards are missed during the monitoring process.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The object of the present invention is to provide a method for enhancing sample target data for transient electromagnetic surface detection, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for enhancing sample target data for transient electromagnetic surface detection, comprising the following steps:

[0008] Step 1: The target detection area is evenly divided into several characteristic unit grids, each characteristic unit grid is marked with a unique identifier, the center point of each characteristic unit grid is used as a monitoring point, and transient electromagnetic signal data at each monitoring point is collected as additional attributes of the corresponding characteristic unit grid. The transient electromagnetic signal data includes: induced voltage time series, three-component magnetic field strength, excitation-reception delay, and background noise level;

[0009] Step 2: When collecting transient electromagnetic signal data at each monitoring point, determine whether the signal value at each monitoring point is missing. If missing, define this monitoring point as a data missing point and fill the data of this data missing point based on the spatial distance attenuation law to complete the data filling of the entire target detection area:

[0010] Step 3: Identify the noise interference monitoring point through a multi-dimensional anomaly detection algorithm, set the neighborhood search radius of the noise interference monitoring point, and correct the transient electromagnetic signal of the noise interference monitoring point according to the signal value of the non-noise interference monitoring point within the neighborhood search radius;

[0011] Step 4: Build a BIM model of the target detection area, align the monitoring points with the model, and update the data filled in step 2 and the correction data of the noise interference monitoring points in step 3 to the corresponding monitoring points in real time. Define a dynamic risk threshold. If the transient electromagnetic signal at the monitoring point exceeds the dynamic risk threshold, the unexploded ordnance warning signal is triggered.

[0012] Furthermore, the target detection area is evenly divided into several rectangular feature unit grids according to the area, and the three-dimensional coordinates of the monitoring point at each feature unit grid are recorded, expressed as (X, Y, Z);

[0013] Label each grid as ; ; is the index of the row number, is the index of the column number, is the total number of rows, is the total number of columns;

[0014] A virtual monitoring point is set up at the center point of each characteristic unit grid, and a transient electromagnetic detection system is used to collect the following transient electromagnetic signal data: induced voltage time series, three-component magnetic field strength, excitation-reception delay and background noise level. A mapping relationship is established between the collected transient electromagnetic signal parameters and the corresponding characteristic unit grid, and the data is stored as a structured data packet with a timestamp.

[0015] Furthermore, when a monitoring point meets any of the following conditions, it is determined that the monitoring point is a data missing point:

[0016] In the transient electromagnetic signal data of the monitoring point, the values ​​of the induced voltage time series, three-component magnetic field strength, excitation-reception delay, or background noise level are non-numeric, empty, or invalid placeholders;

[0017] Any value in the transient electromagnetic signal data of the monitoring point exceeds the effective dynamic range of the instrument calibration;

[0018] The structured data packet of the monitoring point lacks timestamp information, or the deviation between the timestamp of the structured data packet and the system clock exceeds the synchronization tolerance, making it impossible to associate time with other data and making it unusable.

[0019] Furthermore, based on the spatial distance attenuation law, data is filled in for missing data points. The formula is as follows:

[0020] ;

[0021] Indicates the signal value to be filled at the point where data is missing, Indicates the The signal value of a normal monitoring point, is the number of normal monitoring points within the observation radius of the data missing point, is the index of the normal monitoring point within the observation radius of the data missing point, Indicates that the data missing point is The distance between normal monitoring points, is the standard deviation of the Gaussian function, which is used to control the degree of weight attenuation. is a natural constant;

[0022] The observation radius is determined by statistically analyzing the spatial distribution of normal monitoring points, combining the standard deviation and correlation evaluation of signal value changes, to determine the minimum distance that achieves the best filling effect, which is the observation radius;

[0023] Sure The formula is as follows:

[0024] ;

[0025] Where, are the three-dimensional coordinates of the data missing point, For the The three-dimensional coordinates of a normal monitoring point.

[0026] Furthermore, a multi-dimensional anomaly detection algorithm is used to identify noise interference monitoring points. The specific logic is as follows:

[0027] A four-dimensional feature space consisting of time, frequency, space, and physical constraint domains is constructed. The fitting residuals of the induced voltage attenuation curve are analyzed in the time domain. When the sum of squared residuals exceeds three standard deviations of the value predicted by the theoretical attenuation model, a primary anomaly flag is generated.

[0028] The wavelet packet decomposition of the induced voltage time series in the frequency domain is performed to extract the energy ratio characteristics of each sub-band, and the matching degree is analyzed with the typical frequency response characteristic library of unexploded ordnance. The monitoring point generates a secondary abnormal mark;

[0029] Calculate the signal of the monitoring point and the 8-neighborhood in the spatial domain Spatial autocorrelation index: when the index value is less than 0.2, a level 3 anomaly mark is generated;

[0030] Verify in the physical constraint domain whether the signal value at each monitoring point conforms to the discrete solution of Maxwell's equations, and generate a four-level abnormality mark for the monitoring point that violates the basic laws of the electromagnetic field;

[0031] When a monitoring point satisfies the abnormal marking conditions of at least three dimensions at the same time, it is determined to be a noise interference monitoring point.

[0032] Furthermore, the induced voltage time series in the collected transient electromagnetic signal is decomposed by wavelet packet in the frequency domain. The specific logic is as follows:

[0033] The induced voltage timing sequence Perform wavelet packet decomposition, where Daubechies wavelet is selected as the wavelet basis for transformation. The mathematical expression of wavelet packet decomposition is as follows:

[0034] ;

[0035] Where, After decomposition, Tier The coefficients of the subbands, is the induced voltage timing sequence, is the wavelet packet basis function;

[0036] The energy of each subband is calculated according to the following formula:

[0037] ;

[0038] Where, For the Tier The energy brought by the child, Indicates the Tier The first sub-band The square of the absolute value of the wavelet packet coefficients, Indicates the accumulation of all sample points. Indicates that in this subband The index of the next wavelet packet coefficient;

[0039] The energy ratio feature of each subband is calculated based on the following formula:

[0040] ;

[0041] Where, For the Tier The energy ratio characteristics of the sub-bands, the denominator Indicates the accumulation of energy of all sub-bands at all levels, represents the index of the level, Indicates the index of the subband;

[0042] Based on the cosine similarity method, the extracted energy ratio feature Typical frequency response characteristics library of unexploded ordnance Calculate the matching degree:

[0043] ;

[0044] Where, Indicates the matching degree, For the Tier The energy ratio characteristics of the sub-bands, Indicates the corresponding eigenvalue in the typical frequency response feature library of unexploded ordnance;

[0045] If the matching If it is lower than 0.6, the monitoring point is marked as a secondary anomaly, indicating the presence of noise interference.

[0046] Furthermore, for the noise interference monitoring point, a neighborhood search radius is set, and the transient electromagnetic signal of the noise interference monitoring point is corrected according to the signal value of the non-noise interference monitoring point within the neighborhood search radius, specifically:

[0047] Set the neighborhood search radius, obtain all non-noise interference monitoring points within the neighborhood search radius, collect transient electromagnetic signal data of non-noise interference monitoring points, construct correction sample signal values, and correct the signal values ​​of noise interference monitoring points according to the mean and standard deviation of the correction sample signal values. If the original signal value exceeds the mean of the correction sample signal value, the signal value of the noise interference monitoring point is corrected. times the standard deviation range, it is directly replaced by the mean of the corrected sample signal value; if the original signal value is within the mean of the corrected sample signal value When the mean value is within the range of times the standard deviation, weighted fusion is used to correct the original signal value, where the weight of the corrected sample mean is set to 0.6 and the weight of the original signal value is set to 0.4.

[0048] Furthermore, a BIM model of the target detection area is constructed, and the three-dimensional coordinates of the feature unit grid are accurately aligned with the BIM model space coordinate system, and the alignment error is controlled within the grid side length. Map the data filled in step 2 and the data corrected in step 3 to the corresponding monitoring points in the order of timestamps, and store the following attribute groups at each monitoring point: the corrected induced voltage time series, the three-component magnetic field strength, the excitation-reception delay, and the background noise spectrum characteristics;

[0049] The dynamic risk threshold is the mean value of the induced voltage time series, the mean value of the three-component magnetic field intensity, the mean value of the excitation-reception time delay and the background noise level of the historical unexploded ordnance sample library;

[0050] If the transient electromagnetic signal at a monitoring point exceeds its corresponding dynamic risk threshold, an unexploded ordnance warning signal is triggered, and the point is identified as a potential unexploded ordnance target. At the same time, the characteristic unit grid corresponding to the monitoring point is marked as a warning area in the BIM model.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] This method, through meticulous segmentation and data infill of the target detection area, combined with a multi-dimensional anomaly detection algorithm, effectively identifies and corrects data from monitoring points subject to noise interference. By constructing a BIM model, updating data in real time, and dynamically adjusting risk thresholds, this method not only improves data integrity and reliability but also provides timely early warning of potential unexploded ordnance. This solution effectively enhances the application value and practical effectiveness of transient electromagnetic surface detection technology, ensuring the safety and reliability of detection activities. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0054] Figure 2 For the The signal value of a normal monitoring point is Scatter plot of the relationship between values;

[0055] Figure 3 The numerator part in the matching calculation formula is Spline curve graph. DETAILED DESCRIPTION

[0056] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0057] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0058] Example:

[0059] See also Figure 1 , the present invention provides a technical solution:

[0060] A method for enhancing sample target data for transient electromagnetic surface detection, comprising the following steps:

[0061] Step 1: The target detection area is evenly divided into several characteristic unit grids, each characteristic unit grid is marked with a unique identifier, the center point of each characteristic unit grid is used as a monitoring point, and transient electromagnetic signal data at each monitoring point is collected as additional attributes of the corresponding characteristic unit grid. The transient electromagnetic signal data includes: induced voltage time series, three-component magnetic field strength, excitation-reception delay, and background noise level;

[0062] In this embodiment, the induced voltage timing sequence is acquired using a highly sensitive voltage sensor, and the three-component magnetic field strength is measured using a three-dimensional magnetic field sensor. This sensor can simultaneously capture changes in magnetic field strength in three orthogonal directions, ensuring comprehensive magnetic field information. The measurement of the excitation-reception delay requires the use of a time-synchronized signal generator and receiver. These devices can precisely control the transmission and reception times of the signal, thereby calculating the signal propagation delay. Furthermore, the background noise level is monitored using a noise analyzer that can capture and analyze the intensity of ambient noise in different frequency bands to determine its impact on signal quality.

[0063] The target detection area is evenly divided into several rectangular feature unit grids according to the area, and the three-dimensional coordinates of the monitoring point at each feature unit grid are recorded, expressed as (X, Y, Z);

[0064] Label each grid as ; ; is the index of the row number, is the index of the column number, is the total number of rows, is the total number of columns;

[0065] A virtual monitoring point is set up at the center point of each characteristic unit grid, and a transient electromagnetic detection system is used to collect the following transient electromagnetic signal data: induced voltage time series, three-component magnetic field strength, excitation-reception delay and background noise level. A mapping relationship is established between the collected transient electromagnetic signal parameters and the corresponding characteristic unit grid, and the data is stored as a structured data packet with a timestamp.

[0066] The advantage of step 1 is that it evenly divides the target detection area into a grid of characteristic cells and labels each grid with a unique identifier, ensuring systematic and structured data collection. By setting a monitoring point at the center of each grid, comprehensive information about transient electromagnetic signals can be effectively acquired, including multiple parameters such as induced voltage and three-component magnetic field strength. This approach not only increases the density of monitoring point acquisition but also facilitates subsequent data analysis and processing, ensuring data integrity and consistency.

[0067] Compared to existing technologies, this step overcomes the problems of uneven detection area division or irrational monitoring point distribution in the past by evenly dividing the feature unit grid. Existing methods may result in missing signal data or insufficient acquisition in certain areas. However, this solution, through systematic grid division, ensures that every area can be effectively monitored, reduces the risk of signal loss, and improves the reliability and accuracy of detection. In the present invention, the implementation of step 1 lays the foundation for subsequent data filling, anomaly detection, and analysis. Because it ensures the systematization and structuring of data, subsequent steps, such as data filling and noise interference correction, can be carried out with the support of high-quality data, thereby improving the effectiveness and efficiency of the overall solution. In addition, the clear division of monitoring points also makes the results more traceable and verifiable, providing a reliable basis for the final unexploded ordnance early warning.

[0068] Step 2: When collecting transient electromagnetic signal data at each monitoring point, determine whether the signal value at each monitoring point is missing. If missing, define this monitoring point as a data missing point and fill the data of this data missing point based on the spatial distance attenuation law to complete the data filling of the entire target detection area:

[0069] In this embodiment, when a monitoring point meets any of the following conditions, the monitoring point is determined to be a data missing point:

[0070] In the transient electromagnetic signal data of the monitoring point, the values ​​of the induced voltage time series, three-component magnetic field strength, excitation-reception delay, or background noise level are non-numeric, empty, or invalid placeholders;

[0071] Any value in the transient electromagnetic signal data of the monitoring point exceeds the effective dynamic range of the instrument calibration;

[0072] The structured data packet of the monitoring point lacks timestamp information, or the deviation between the timestamp of the structured data packet and the system clock exceeds the synchronization tolerance, making it impossible to associate time with other data and making it unusable.

[0073] When transient electromagnetic signal data at a monitoring point meets any of the above conditions, it is considered a missing data point because the validity and reliability of the data are severely affected. First, when the values ​​of the induced voltage time series, three-component magnetic field strength, excitation-reception delay, or background noise level are non-numeric, null, or invalid placeholders, this indicates that an error or failure occurred during the data acquisition process, making it impossible to provide true measurement information and, therefore, unusable for subsequent analysis.

[0074] Secondly, if any value in the monitoring point data exceeds the effective dynamic range of the instrument calibration, the data may be an outlier due to sensor saturation or failure. Such data usually has no physical meaning and may cause distortion of the analysis results. Therefore, such data should also be considered invalid.

[0075] Finally, the lack of timestamp information or the deviation of the timestamp from the system clock exceeding the synchronization tolerance means that the data at that monitoring point cannot be effectively associated with other data. The lack of time information hinders the temporal analysis and comparison of the data, making it impossible to integrate and interpret this data with data from other monitoring points in the time dimension. Therefore, to ensure the accuracy of data processing and the reliability of results, data in these cases must be identified as missing points to avoid introducing misleading results in subsequent analysis.

[0076] Based on the spatial distance attenuation law, data is filled in for missing data points. The formula is as follows:

[0077] ;

[0078] Indicates the signal value to be filled at the point where data is missing, Indicates the The signal value of a normal monitoring point, is the number of normal monitoring points within the observation radius of the data missing point, is the index of the normal monitoring point within the observation radius of the data missing point, Indicates that the data missing point is The distance between normal monitoring points, is the standard deviation of the Gaussian function, which is used to control the degree of weight attenuation. is a natural constant; the observation radius is 5 times the side length of the characteristic unit grid, and it is ensured that at least 8 normal monitoring points are included.

[0079] In the above formula, the dependent variable It reflects the signal value at the data missing point, calculated by considering the signal data of the surrounding normal monitoring points and the distance attenuation relationship. This represents the expected signal strength at the missing data point, calculated by weighted averaging the signal characteristics of neighboring monitoring points. This filling method leverages spatial similarity and effectively reduces analytical errors caused by missing data. Compared to simple interpolation or constant filling, filling based on distance decay more accurately reflects the actual conditions at the monitoring point, thereby improving data integrity and reliability and ensuring the accuracy of subsequent analysis.

[0080] Independent variable With dependent variable The relationship between is based on the logic of spatial distance and signal similarity, specifically: It reflects the actual signal strength of a normal monitoring point at a specific location. These signal strengths are affected by the surrounding environment and target characteristics and usually have spatial correlation. As the value of increases, the influence of the signal will be weakened, that is, the contribution of the farther monitoring points to the fill signal will be reduced. , this relationship is quantified. Therefore, the independent variable Changes in the dependent variable will directly affect The calculated value of . When increasing, Keep it between 0 and 1, so that The calculation result of also increases accordingly; on the contrary, when When it decreases, the corresponding will also decrease as the contribution of lower normal monitoring point signal values ​​to the fill signal decreases.

[0081] The formal rationality of this formula is reflected in its effective combination of the relationship between signal value and spatial distance, which reflects the influence of normal monitoring points on data missing points in a weighted attenuation manner. , the closer the monitoring point is, the greater its contribution to filling the missing points, and as the distance increases, the weight gradually decreases, reflecting the natural law of signal attenuation in space. This design can effectively avoid the interference of irrelevant signals far away from the monitoring point on the filling result, and improve the accuracy and reliability of filling. In addition, the standard deviation The introduction of can flexibly adjust the speed of weight decay, so that the filling process can be optimized according to the actual situation, further enhancing the rationality and effectiveness of the application of the formula in surface detection.

[0082] Table 1: SMIS value statistics

[0083]

[0084] See also Figure 2 , this data analysis includes 20 data, each group contains Signal value of normal monitoring points, number of normal monitoring points, distance to missing points, and filling results The number and distance of normal monitoring points in each group in the data are different, which will directly affect the calculation of the filling results.

[0085] In each data set, it was observed that The signal value of a normal monitoring point With filling results There is a significant correlation between the two. For example, the signal value of group 7 is 18, and the distance is 0.8, and the filling result is as high as 17.86. This shows that monitoring points closer to the missing point contribute significantly to the filling result. Even if their signal values ​​are relatively high, they can still effectively improve the accuracy of the signal value to be filled.

[0086] distance It is a significant factor influencing the filling results. The analysis of group 11 shows that despite a signal value of 25 and a distance of 5, the final filling result was 21.12, demonstrating that the influence of high signal values ​​decreases at longer distances. This confirms the characteristics of the Gaussian decay function: as distance increases, the weight gradually decreases, resulting in a decrease in the filling result.

[0087] Number of normal monitoring points It also significantly affects the filling effect. For example, group 4 has 5 normal monitoring points with a distance of 3. The value of group 2 is 10.85, while group 2 has a higher signal value of 15, but the number of normal monitoring points is only 4, and its filling result is 14.11. This suggests that increasing the number of effective monitoring points can compensate for the negative impact of low signal values ​​on filling results to a certain extent.

[0088] The filling method based on the spatial distance attenuation law shows good results in processing missing data points. By using the signal value and distance information of normal monitoring points, the missing signal value can be effectively calculated. The analysis results show that the interaction between signal value, distance, and the number of normal monitoring points is crucial for understanding the filling process and its accuracy. This method provides a scientific basis for subsequent data analysis and decision-making, and has good application prospects.

[0089] The observation radius is determined by statistically analyzing the spatial distribution of normal monitoring points, combining the standard deviation and correlation evaluation of signal value changes, to determine the minimum distance that achieves the best filling effect, which is the observation radius;

[0090] Sure The formula is as follows:

[0091] ;

[0092] Where, are the three-dimensional coordinates of the data missing point, For the The three-dimensional coordinates of a normal monitoring point;

[0093] According to the signal value to be filled in the missing data point , fill in the missing parts with data, thereby completing the data filling of the entire target detection area.

[0094] The advantage of step 2 is that it systematically determines whether signal values ​​at monitoring points are missing, thereby ensuring data integrity and accuracy. By defining missing monitoring points as data-missing points and filling in data based on the spatial distance decay principle, this method effectively repairs data gaps and ensures that the signal data used in subsequent analysis is complete. This data-filling method, by considering spatial relationships, makes the supplemented data more reliable, thereby improving the credibility of the overall monitoring results.

[0095] Compared with existing technologies, this step reduces the interference of unqualified data on analysis results by using clear missing data judgment criteria, such as signal values ​​that are non-numeric, exceed the dynamic range, or lack timestamps. Existing methods may lack effective missing data processing mechanisms, resulting in the use of unreliable data in the analysis. However, this step ensures data integrity through scientific filling methods, significantly improving data quality and the accuracy of subsequent analysis. In the present invention, the implementation of step 2 lays the foundation for subsequent anomaly detection and risk assessment. By filling in missing data, the continuity and consistency of monitoring data are ensured, allowing subsequent steps, such as noise interference detection and the setting of dynamic risk thresholds, to be carried out with the support of high-quality data. This not only improves the accuracy and reliability of monitoring, but also provides a solid data foundation for the timely triggering of unexploded ordnance warning signals, enhancing the effectiveness and safety of the overall detection system.

[0096] Step 3: Identify the noise interference monitoring point through a multi-dimensional anomaly detection algorithm, set the neighborhood search radius of the noise interference monitoring point, and correct the transient electromagnetic signal of the noise interference monitoring point according to the signal value of the non-noise interference monitoring point within the neighborhood search radius;

[0097] In this embodiment, a multi-dimensional anomaly detection algorithm is used to identify noise interference monitoring points. The specific logic is as follows:

[0098] A four-dimensional feature space consisting of time, frequency, space, and physical constraint domains is constructed. The fitting residuals of the induced voltage attenuation curve are analyzed in the time domain. When the sum of squared residuals exceeds three standard deviations of the value predicted by the theoretical attenuation model, a primary anomaly flag is generated.

[0099] The wavelet packet decomposition of the induced voltage time series in the frequency domain is performed to extract the energy ratio characteristics of each sub-band, and the matching degree is analyzed with the typical frequency response characteristic library of unexploded ordnance. The monitoring point generates a secondary abnormal mark;

[0100] Calculate the signal of the monitoring point and the 8-neighborhood in the spatial domain Spatial autocorrelation index: when the index value is less than 0.2, a level 3 anomaly mark is generated;

[0101] Verify in the physical constraint domain whether the signal value at each monitoring point conforms to the discrete solution of Maxwell's equations, and generate a four-level abnormality mark for the monitoring point that violates the basic laws of the electromagnetic field;

[0102] When a monitoring point satisfies the abnormal marking conditions of at least three dimensions at the same time, it is determined to be a noise interference monitoring point.

[0103] In the constructed four-dimensional feature space, in the time domain, the residuals of the induced voltage attenuation curve are analyzed. If the sum of squared residuals exceeds three standard deviations of the theoretical attenuation model's predicted value, a primary anomaly flag is generated, indicating a significant deviation in the voltage signal at the monitoring point, possibly affected by noise or interference. Secondly, wavelet packet decomposition is performed in the frequency domain to extract the energy ratio features of each subband. This is then compared with a typical frequency response feature library of unexploded ordnance for matching. If the matching degree is less than 0.6, a secondary anomaly flag is generated, indicating that the signal characteristics of the monitoring point differ significantly from the normal state and may be affected by abnormal noise. In the spatial domain, the Moran's I spatial autocorrelation index is calculated between the monitoring point and its eight neighboring signals. When the index value is less than 0.2, a tertiary anomaly flag is generated, indicating that the signal at the monitoring point is isolated from the surrounding environment, possibly due to the influence of environmental noise. Finally, in the physical constraint domain, the signal values ​​of each monitoring point are verified to ensure that they conform to the discrete solution of Maxwell's equations. If a monitoring point is found to violate the basic laws of the electromagnetic field, a tertiary anomaly flag is generated, indicating that the signal is severely interfered with or affected by false positives. When a monitoring point meets the abnormal marking conditions of at least three dimensions at the same time, it is determined to be a noise interference monitoring point, indicating that the point may be affected by external noise interference, electromagnetic interference or signal distortion, resulting in unreliable data. The wavelet packet decomposition of the induced voltage time series in the collected transient electromagnetic signal is performed in the frequency domain. The specific logic is as follows:

[0104] The induced voltage timing sequence Perform wavelet packet decomposition, where Daubechies wavelet is selected as the wavelet basis for transformation. The mathematical expression of wavelet packet decomposition is as follows:

[0105] ;

[0106] Where, After decomposition, Tier The coefficient of each sub-band reflects the energy distribution of the signal within the frequency range of the sub-band. is the induced voltage timing sequence, is the wavelet packet basis function;

[0107] Through this process, the signal is decomposed into multiple frequency bands, so that each sub-band provides information about the signal's energy within a specific frequency range, laying the foundation for subsequent feature extraction and analysis. This approach effectively enhances the understanding of complex signals and can better identify potential patterns or anomalies.

[0108] The energy of each subband is calculated according to the following formula:

[0109] ;

[0110] Where, For the Tier The energy brought by the child, Indicates the Tier The first sub-band The square of the absolute value of the wavelet packet coefficients, Indicates the accumulation of all sample points. Indicates that in this subband The index of the next wavelet packet coefficient;

[0111] By accumulating the square of the absolute value of all wavelet packet coefficients in the sub-band, the total energy of the signal within the frequency range of the sub-band can be reflected. This calculation method effectively captures the energy distribution of the signal in a specific frequency band and helps analyze the frequency characteristics and dynamic changes of the signal. Specifically, Represents the The energy of each wavelet packet coefficient reflects the importance of the frequency component to the overall signal. By summing all sample points, the energy characteristics of the sub-band can be comprehensively evaluated, providing an important basis for subsequent data analysis and feature extraction.

[0112] The energy ratio feature of each subband is calculated based on the following formula:

[0113] ;

[0114] Where, For the Tier The energy ratio characteristics of the sub-bands, the denominator Indicates the accumulation of energy of all sub-bands at all levels, represents the index of the level, Indicates the index of the subband;

[0115] The normalized approach described above allows the energy ratio of each subband to reflect its relative importance within the entire signal, helping analysts identify which frequency bands contribute most significantly to the overall signal. This approach effectively eliminates the influence of energy differences between subbands, enhancing understanding of signal characteristics and providing a more accurate basis for subsequent pattern recognition and feature extraction.

[0116] Based on the cosine similarity method, the extracted energy ratio feature Typical frequency response characteristics library of unexploded ordnance Calculate the matching degree:

[0117] ;

[0118] Where, Indicates the matching degree, For the Tier The energy ratio characteristics of the sub-bands, Indicates the corresponding eigenvalue in the typical frequency response feature library of unexploded ordnance;

[0119] Match It reflects the consistency between the extracted features and the reference features. For the Tier The energy ratio characteristics of the sub-bands are is the corresponding characteristic value in the typical frequency response characteristic library of unexploded ordnance. The dot product between the two eigenvectors is calculated to represent their similarity in the feature space; the denominator is normalized by calculating their respective norms to ensure that the matching value is between 0 and 1. A value close to 1 indicates a high degree of similarity between the extracted energy ratio feature and features in the typical frequency response signature library of unexploded ordnance. This suggests that the signal may share the frequency response characteristics of unexploded ordnance, suggesting that the source of the signal may be related to unexploded ordnance. This facilitates comparison between different features. This method can effectively identify potential connections between signals and unexploded ordnance, providing important support for safety assessments and decision-making.

[0120] If the matching If it is lower than 0.6, the monitoring point is marked as a secondary anomaly, indicating the presence of noise interference.

[0121] Table 2: Matching SQ Statistics

[0122]

[0123] See also Figure 3 It should be noted that the dot product of the two eigenvectors in the table corresponds to the numerator of the formula In this data analysis, the cosine similarity was calculated to evaluate the matching degree between the energy ratio feature and the typical frequency response feature library of unexploded ordnance, and the corresponding matching degree data was obtained. Increase in matching This relationship indicates that when the similarity between the energy ratio feature and the typical frequency response feature increases, the matching degree also increases.

[0124] Specifically, the initial matching degree is When the similarity is 0.707, the matching degree gradually increases to 0.8, and 0.991, and finally The maximum matching degree of 1 is achieved when . This result verifies the effectiveness of the model in feature matching and shows that the energy ratio feature can better reflect the frequency response characteristics of unexploded ordnance.

[0125] The above analysis results provide a solid foundation for further research and application, and emphasize that the ability to identify unexploded ordnance characteristics can be enhanced through effective feature extraction and matching methods, which is of great practical significance for improving safety detection and early warning mechanisms.

[0126] For the noise interference monitoring point, a neighborhood search radius is set, and the transient electromagnetic signal of the noise interference monitoring point is corrected according to the signal value of the non-noise interference monitoring point within the neighborhood search radius. Specifically:

[0127] Set the neighborhood search radius, obtain all non-noise interference monitoring points within the neighborhood search radius, collect transient electromagnetic signal data of non-noise interference monitoring points, construct correction sample signal values, and correct the signal values ​​of noise interference monitoring points according to the mean and standard deviation of the correction sample signal values. If the original signal value exceeds the mean of the correction sample signal value, the signal value of the noise interference monitoring point is corrected. times the standard deviation range, it is directly replaced by the mean of the corrected sample signal value; if the original signal value is within the mean of the corrected sample signal value If the signal is within the range of 1 / 3 standard deviation, weighted fusion is used to correct the original signal value, where the weight of the corrected sample mean is set to 0.6 and the weight of the original signal value is set to 0.4. The neighborhood search radius is 3 times the side length of the current feature unit grid, and it is ensured to include at least 8 adjacent normal monitoring points and not exceed the maximum effective detection range of 20 meters.

[0128] The advantage of step 3 is that the multi-dimensional anomaly detection algorithm effectively identifies noise-interfering monitoring points, thereby improving data quality and reliability. By setting a neighborhood search radius and correcting the affected monitoring points based on the signal values ​​of non-noise-interfering monitoring points within the neighborhood, the effects of abnormal interference on the signal can be accurately eliminated. This approach not only improves the accuracy of monitoring results but also makes subsequent data analysis more effective, making it a key step in ensuring the overall performance of the monitoring system.

[0129] Compared to existing technologies, this step provides a more comprehensive and systematic anomaly detection mechanism by constructing a four-dimensional feature space encompassing the time, frequency, spatial, and physical constraint domains. Existing methods often rely on a single indicator or simple rules to determine noise interference, which can lead to misjudgments or missed detections. However, the multidimensional analysis in this step integrates diverse information, significantly improving the accuracy and reliability of anomaly detection and ensuring that the processed data is more authentic and effective. In the present invention, the implementation of step 3 not only strengthens the detection and correction of noise interference but also provides a solid foundation for the accurate analysis of monitoring data. By ensuring the reliability of monitoring point data, subsequent steps such as the setting of dynamic risk thresholds and the triggering of unexploded ordnance warnings can be performed with higher data quality, thereby enhancing the effectiveness and safety of the overall solution. This process ensures that the monitoring system can promptly and accurately identify potential safety risks, improving the efficiency and accuracy of emergency response.

[0130] Step 4: Build a BIM model of the target detection area, align the monitoring points with the model, and update the corresponding monitoring points in real time with the data filled in step 2 and the correction data of the noise interference monitoring points in step 3. Define a dynamic risk threshold. If the transient electromagnetic signal at the monitoring point exceeds the dynamic risk threshold, the unexploded ordnance warning signal is triggered.

[0131] In this embodiment, a BIM model of the target detection area is constructed, and the three-dimensional coordinates of the feature unit grid are accurately aligned with the BIM model space coordinate system, and the alignment error is controlled within the grid side length. Map the data filled in step 2 and the data corrected in step 3 to the corresponding monitoring points in the order of timestamps, and store the following attribute groups at each monitoring point: the corrected induced voltage time series, the three-component magnetic field strength, the excitation-reception delay, and the background noise spectrum characteristics;

[0132] The dynamic risk threshold is the mean value of the induced voltage time series, the mean value of the three-component magnetic field intensity, the mean value of the excitation-reception time delay and the background noise level of the historical unexploded ordnance sample library;

[0133] If the transient electromagnetic signal at a monitoring point exceeds its corresponding dynamic risk threshold, an unexploded ordnance warning signal is triggered, and the point is identified as a potential unexploded ordnance target. At the same time, the characteristic unit grid corresponding to the monitoring point is marked as a warning area in the BIM model.

[0134] The advantage of step 4 lies in precisely registering monitoring points with a BIM (Building Information Model) model of the target detection area, ensuring the spatial accuracy of the monitoring data. This approach allows for effective visualization and analysis of monitoring data in three dimensions, enhancing its comprehensibility and practicality. Furthermore, the definition of dynamic risk thresholds and real-time updating of monitoring point data enable the system to respond quickly and provide timely warnings of potential unexploded ordnance risks, thereby enhancing the safety and effectiveness of detection.

[0135] Compared with the existing technology, step 4 provides a more systematic and intelligent data processing method by introducing the construction of BIM models and the setting of dynamic risk thresholds. Existing methods often use static monitoring and analysis methods, which make it difficult to respond to environmental changes and data fluctuations in real time. This step, by updating the data of monitoring points in real time and setting dynamic thresholds, enables the monitoring system to adapt to complex detection environments, significantly improves the timeliness and accuracy of early warnings, and reduces the occurrence of potential risks. In the present invention, the implementation of step 4 provides a comprehensive framework for the entire data enhancement processing method, ensuring the accuracy, real-time nature and practicality of the monitoring data. By synchronizing the filled data and the corrected data to the BIM model, the overall solution can achieve more efficient management and response. This process not only improves the scientific nature and systematic nature of data processing, but also provides a solid data foundation for subsequent decision-making, thereby greatly enhancing the detection and early warning capabilities of unexploded ordnance and ensuring public safety.

[0136] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0137] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0138] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0139] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A sample target data enhancement processing method for transient electromagnetic surface detection, characterized in that: The specific steps include: Step 1: The target detection area is evenly divided into several characteristic unit grids, each characteristic unit grid is marked with a unique identifier, the center point of each characteristic unit grid is used as a monitoring point, and transient electromagnetic signal data at each monitoring point is collected as additional attributes of the corresponding characteristic unit grid. The transient electromagnetic signal data includes: induced voltage time series, three-component magnetic field strength, excitation-reception delay, and background noise level; Step 2: When collecting transient electromagnetic signal data at each monitoring point, determine whether the signal value at each monitoring point is missing. If missing, define this monitoring point as a data missing point and fill the data of this data missing point based on the spatial distance attenuation law to complete the data filling of the entire target detection area: Step 3: Identify the noise interference monitoring point through a multi-dimensional anomaly detection algorithm, set the neighborhood search radius of the noise interference monitoring point, and correct the transient electromagnetic signal of the noise interference monitoring point according to the signal value of the non-noise interference monitoring point within the neighborhood search radius; Step 4: Build a BIM model of the target detection area, align the monitoring points with the model, and update the data filled in step 2 and the correction data of the noise interference monitoring points in step 3 to the corresponding monitoring points in real time. Define a dynamic risk threshold. If the transient electromagnetic signal at the monitoring point exceeds the dynamic risk threshold, the unexploded ordnance warning signal is triggered.

2. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 1, characterized in that: The target detection area is evenly divided into several rectangular feature unit grids according to the area, and the three-dimensional coordinates of the monitoring point at each feature unit grid are recorded, expressed as (X, Y, Z); Label each grid as ; ; is the index of the row number, is the index of the column number, is the total number of rows, is the total number of columns; A virtual monitoring point is set up at the center point of each characteristic unit grid, and a transient electromagnetic detection system is used to collect the following transient electromagnetic signal data: induced voltage time series, three-component magnetic field strength, excitation-reception delay and background noise level. A mapping relationship is established between the collected transient electromagnetic signal parameters and the corresponding characteristic unit grid, and the data is stored as a structured data packet with a timestamp.

3. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 2, characterized in that: When a monitoring point meets any of the following conditions, it is determined to be a data missing point: In the transient electromagnetic signal data of the monitoring point, the values ​​of the induced voltage time series, three-component magnetic field strength, excitation-reception delay, or background noise level are non-numeric, empty, or invalid placeholders; Any value in the transient electromagnetic signal data of the monitoring point exceeds the effective dynamic range of the instrument calibration; The structured data packet of the monitoring point lacks timestamp information, or the deviation between the timestamp of the structured data packet and the system clock exceeds the synchronization tolerance, making it impossible to associate time with other data and making it unusable.

4. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 1, characterized in that: Based on the spatial distance attenuation law, data is filled in for missing data points. The formula is as follows: ； Indicates the signal value to be filled at the point where data is missing, Indicates the The signal value of a normal monitoring point, is the number of normal monitoring points within the observation radius of the data missing point, is the index of the normal monitoring point within the observation radius of the data missing point, Indicates that the data missing point is The distance between normal monitoring points, is the standard deviation of the Gaussian function, which is used to control the degree of weight attenuation. is a natural constant; The observation radius is determined by statistically analyzing the spatial distribution of normal monitoring points, combining the standard deviation and correlation evaluation of signal value changes, to determine the minimum distance that achieves the best filling effect, which is the observation radius; Sure The formula is as follows: ； Where, are the three-dimensional coordinates of the data missing point, For the The three-dimensional coordinates of a normal monitoring point.

5. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 1, characterized in that: The multi-dimensional anomaly detection algorithm is used to identify noise interference monitoring points. The specific logic is as follows: A four-dimensional feature space consisting of time, frequency, space, and physical constraint domains is constructed. The fitting residuals of the induced voltage attenuation curve are analyzed in the time domain. When the sum of squared residuals exceeds three standard deviations of the value predicted by the theoretical attenuation model, a primary anomaly flag is generated. The wavelet packet decomposition of the induced voltage time series in the frequency domain is performed to extract the energy ratio characteristics of each sub-band, and the matching degree is analyzed with the typical frequency response characteristic library of unexploded ordnance. The monitoring point generates a secondary abnormal mark; Calculate the signal of the monitoring point and the 8-neighborhood in the spatial domain Spatial autocorrelation index: when the index value is less than 0.2, a level 3 anomaly mark is generated; Verify in the physical constraint domain whether the signal value at each monitoring point conforms to the discrete solution of Maxwell's equations, and generate a four-level abnormality mark for the monitoring point that violates the basic laws of the electromagnetic field; When a monitoring point satisfies the abnormal marking conditions of at least three dimensions at the same time, it is determined to be a noise interference monitoring point.

6. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 5, characterized in that: The wavelet packet decomposition of the induced voltage time series in the acquired transient electromagnetic signal is performed in the frequency domain. The specific logic is as follows: The induced voltage timing sequence Perform wavelet packet decomposition, where Daubechies wavelet is selected as the wavelet basis for transformation. The mathematical expression of wavelet packet decomposition is as follows: ； Where, Indicates that after decomposition Tier The coefficients of the subbands, is the induced voltage timing sequence, is the wavelet packet basis function; The energy of each subband is calculated according to the following formula: ； Where, For the Tier The energy brought by the child, Indicates the Tier The first sub-band The square of the absolute value of the wavelet packet coefficients, Indicates the accumulation of all sample points. Indicates that in this subband The index of the next wavelet packet coefficient; The energy ratio feature of each subband is calculated based on the following formula: ； Where, For the Tier The energy ratio characteristics of the sub-bands, the denominator Indicates the accumulation of energy of all sub-bands at all levels, represents the index of the level, Indicates the index of the subband; Based on the cosine similarity method, the extracted energy ratio feature Typical frequency response feature library of unexploded ordnance Calculate the matching degree: ； Where, Indicates the matching degree, For the Tier The energy ratio characteristics of the sub-bands, Indicates the corresponding eigenvalue in the typical frequency response feature library of unexploded ordnance; If the matching If it is lower than 0.6, the monitoring point is marked as a secondary anomaly, indicating the presence of noise interference.

7. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 5, characterized in that: For the noise interference monitoring point, a neighborhood search radius is set, and the transient electromagnetic signal of the noise interference monitoring point is corrected according to the signal value of the non-noise interference monitoring point within the neighborhood search radius. Specifically: Set the neighborhood search radius, obtain all non-noise interference monitoring points within the neighborhood search radius, collect transient electromagnetic signal data of non-noise interference monitoring points, construct correction sample signal values, and correct the signal values ​​of noise interference monitoring points according to the mean and standard deviation of the correction sample signal values. If the original signal value exceeds the mean of the correction sample signal value, the signal value of the noise interference monitoring point is corrected. times the standard deviation range, it is directly replaced by the mean of the corrected sample signal value; if the original signal value is within the mean of the corrected sample signal value When the mean value is within the range of times the standard deviation, weighted fusion is used to correct the original signal value, where the weight of the corrected sample mean is set to 0.6 and the weight of the original signal value is set to 0.

4.

8. The method for enhancing sample target data for transient electromagnetic surface detection according to claim 1, characterized in that: Construct a BIM model of the target detection area, accurately align the three-dimensional coordinates of the feature unit grid with the BIM model space coordinate system, and control the alignment error within the grid side length. Map the data filled in step 2 and the data corrected in step 3 to the corresponding monitoring points in the order of timestamps, and store the following attribute groups at each monitoring point: the corrected induced voltage time series, the three-component magnetic field strength, the excitation-reception delay, and the background noise spectrum characteristics; The dynamic risk threshold is the mean value of the induced voltage time series, the mean value of the three-component magnetic field intensity, the mean value of the excitation-reception time delay and the background noise level of the historical unexploded ordnance sample library; If the transient electromagnetic signal at a monitoring point exceeds its corresponding dynamic risk threshold, an unexploded ordnance warning signal is triggered, and the point is identified as a potential unexploded ordnance target. At the same time, the characteristic unit grid corresponding to the monitoring point is marked as a warning area in the BIM model.

Images