A handheld detection system for spatial safety

Abstract

This invention relates to the field of security detection technology and discloses a handheld detection system for spatial security, comprising: a static fingerprint construction module for establishing a static fingerprint database and a whitelist shielding domain; a spatiotemporal sampling acquisition module for acquiring continuous spatiotemporal sequences via a handheld device; a difference correction analysis module for correcting sampling and constructing a composite difference degree; a candidate clustering screening module for clustering and screening candidate clusters under whitelist constraints; a gradient approximation confirmation module for performing difference gradient approximation and peak region kernel confirmation to obtain stable peak points; and a close-range verification and review module for verification and positioning based on thermal inertia and geomagnetic steady-state characteristics, and combining multi-path verification to output results. This invention achieves high-precision, stable, and unique detection and positioning of potential anomaly sources in complex spatial environments.

Description

Technical Field

[0001] This invention belongs to the field of safety detection technology, specifically relating to a handheld detection system for spatial safety. Background Technology

[0002] With increasingly stringent security requirements in public places and key areas, the rapid detection of potential anomalies and suspicious devices has become a pressing issue. Current detection methods primarily rely on fixed monitoring equipment or single-type sensors, such as infrared detectors, geomagnetic sensors, or video surveillance systems. However, these methods suffer from inflexible deployment, high environmental dependence, and a relatively high false alarm rate. Fixed monitoring equipment in complex spaces is often limited by its field of view and installation location, making full coverage difficult; single sensors are also susceptible to environmental noise, equipment angle, or human interference, leading to unstable detection results.

[0003] Furthermore, traditional methods often fail to integrate multiple passive environmental parameters, making it difficult to comprehensively characterize the features of anomalous targets. Relying solely on temperature is susceptible to seasonal or airflow influences; relying solely on geomagnetic signals may result in distortion due to interference from metal structures or electrical equipment; and electrostatic field and visible light reflectance parameters are prone to instantaneous fluctuations. Due to the lack of multi-parameter joint judgment and dynamic correction mechanisms, existing detection systems struggle to guarantee accuracy and robustness in complex environments. Summary of the Invention

[0004] This invention provides a handheld detection system for spatial safety, which solves the technical problems in related technologies, such as reliance on a single sensor, susceptibility to environmental interference, lack of multi-parameter fusion and verification mechanisms leading to unstable detection results, high false alarm rate, and difficulty in achieving unique location of anomaly sources.

[0005] This invention provides a handheld detection system for spatial safety, comprising:

[0006] The static fingerprint construction module is used to establish a static fingerprint database and whitelist and its shielding domain. The static fingerprint database is based on passive parameters collected according to a preset grid under a no-suspicion state.

[0007] The spatiotemporal sampling acquisition module is used to acquire continuous spatiotemporal sampling sequences through a handheld device, obtain continuous position sequences based on the handheld inspection path, and synchronously collect passive parameters at each position to form a spatiotemporal sampling sequence.

[0008] The difference correction analysis module is used to correct readings and construct composite difference for spatiotemporal sampling sequences. It is based on posture, walking speed and changes in the neighboring environment for correction. The corrected passive parameters are compared with the static fingerprint database benchmark to obtain the composite difference.

[0009] The candidate clustering and filtering module is used to perform candidate clustering and persistence filtering under whitelist constraints. It performs connectivity clustering on positions where the composite difference exceeds a preset threshold outside the shielded domain. It filters based on the drop and area conditions of the composite difference and outputs the centroid, maximum difference point and principal axis direction of the candidate cluster.

[0010] The gradient approximation confirmation module is used to perform walking approximation and peak kernel confirmation guided by differential gradient using a handheld device. It advances according to the spatial upward direction of the composite difference degree and performs circumferential mapping within the candidate cluster to confirm the peak kernel and obtain stable peak point information.

[0011] The close-range verification module is used to perform close-range passive verification and unique positioning based on thermal inertial characteristics and geomagnetic steady-state fluctuation characteristics. It obtains the positioning result through gradient intersection of stable peak points and calls other modules for verification along different paths.

[0012] Furthermore, the whitelist records the location of common sources and the average value of passive parameters;

[0013] The shielding domain is obtained by merging the influence range of the whitelisted objects;

[0014] The passive parameters include: temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance.

[0015] Furthermore, a static fingerprint database and a whitelist and its masking domain are established. The static fingerprint database collects passive parameters according to a preset grid based on a no-suspect status, including:

[0016] Step 11: Clear all mobile electronic devices from the premises until they are free of suspicion, and divide the premises space according to a preset grid.

[0017] Step 12: Sample each grid multiple times to collect passive parameters, and calculate the mean and standard deviation of temperature, geomagnetic amplitude, electrostatic potential and visible light reflectance respectively. Use the mean and standard deviation as the baseline statistical values ​​of the grid point and store them in the static fingerprint database.

[0018] Step 13: Record the location of common sources related to passive parameters and their corresponding baseline statistics to form a whitelist, and generate corresponding shielding areas based on the influence range of each common source. Merge the shielding areas to obtain the shielding domain.

[0019] Furthermore, a continuous spatiotemporal sampling sequence is acquired using a handheld device, a continuous position sequence is obtained based on the handheld inspection path, and passive parameters are simultaneously collected at each position to form a spatiotemporal sampling sequence, including:

[0020] Step 21: Walk along a preset inspection path using a handheld device. The preset inspection path is an S-shaped path that covers the boundary of the site and extends into the interior. Obtain a continuous position sequence by coordinating drift error constraint based on inertial odometry and visual loop closure algorithm.

[0021] Step 22: At each location point in the continuous location sequence, data is collected in a preset time window. Within each time window, temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance are acquired, and the window mean is calculated to form a parameter sequence.

[0022] Step 23: Bind the continuous position sequence to the parameter sequence one by one, and store the corresponding passive parameters with the coordinates of the position points as indexes to form a spatiotemporal sampling sequence.

[0023] Furthermore, the spatiotemporal sampling sequences are readout corrected and a composite dissimilarity is constructed. Correction is performed based on attitude, walking speed, and changes in the nearest neighbor environment. The corrected passive parameters are compared with a static fingerprint database benchmark to obtain the composite dissimilarity, which includes:

[0024] Step 31: Obtain the attitude angle of the handheld device and compensate and correct the visible light reflectance value in the spatiotemporal sampling sequence. The compensation and correction uses the cosine value of the attitude angle as the correction coefficient.

[0025] Step 32: Determine the weight at each location point based on the walking speed. The weight is obtained by calculating the ratio of the walking speed at the current location to the preset reference speed. During the passive parameter correction process at each location point, differential correction is performed by combining the average value of the sampling points within the preset sampling radius centered on that location point.

[0026] Step 33: The corrected passive parameters are compared with the corresponding baseline values ​​in the static fingerprint database. The comparison uses the standard deviation of each passive parameter in the static fingerprint database as a normalization factor, and calculates the composite difference by taking the square root of the sum of the squares of the differences between each passive parameter and the corresponding baseline value.

[0027] Furthermore, the candidate clustering screening module includes:

[0028] Step 41: Outside the shielding domain, filter the locations where the composite difference exceeds a preset difference threshold to obtain a candidate point set. The preset threshold is the sum of the mean of the composite difference and the standard deviation of a preset multiple.

[0029] Step 42: Cluster the candidate point set according to spatial connectivity to obtain a candidate cluster set. The spatial connectivity is determined based on whether the Euclidean distance between candidate points is less than a preset neighborhood radius. Calculate the centroid, maximum difference point, and principal axis direction for each candidate cluster. The principal axis direction is obtained by performing eigenvalue decomposition on the covariance matrix of the candidate cluster point set.

[0030] Step 43: Calculate the persistence of candidate clusters and filter them in combination with the area condition, and output the centroid, maximum difference point and principal axis direction of the candidate clusters that meet the conditions; the persistence is the difference between the maximum value of the composite difference within the candidate cluster and the minimum value of the composite difference at the convex hull boundary of the candidate cluster, and the area condition is obtained by the number of points within the candidate cluster being greater than or equal to a preset threshold.

[0031] Furthermore, the differential gradient-guided walking approximation and peak region kernel confirmation include:

[0032] Step 51: Calculate the upward direction based on the gradient of the composite difference degree within the candidate cluster, select the position direction with the fastest upward movement as the direction of travel, and gradually move along the direction of travel by carrying a handheld device to update the position points and form an approximation path sequence; the upward direction is obtained by differential calculation of the composite difference degree of adjacent position points within the candidate cluster, and the gradual movement adopts an adaptive step size adjusted according to the gradient magnitude of the composite difference degree.

[0033] Step 52: When the approximation path approaches the local peak region of the candidate cluster, sampling is performed around the peak along a closed path to form a surrounding sampling set. The composite difference of the surrounding sampling set is interpolated on the plane and then contour lines are drawn. The boundary of the peak region kernel is determined according to the shape of the contour lines. The closed path is a ring path with the local peak point as the center.

[0034] Step 53: Within the peak region core, determine the point with the largest composite difference as a candidate peak point. By repeatedly performing sampling under different surrounding mapping paths, if the candidate peak point is detected under at least two surrounding mapping paths and its spatial position difference is less than a preset consistency threshold, then the candidate peak point is determined as a stable peak point, and stable peak point information is output.

[0035] Furthermore, in step 52, the boundary of the peak region is determined based on the fact that the rate of change of the area of ​​the contour-closed region is less than a preset rate of change threshold.

[0036] Furthermore, performing close-range passive verification includes:

[0037] Step 61: Collect continuous temperature data at stable peak points to form a temperature sampling sequence, and calculate the rate of temperature change over time by the difference between adjacent sampling points. When the rate of temperature change is lower than a preset thermal inertia threshold, it is determined that the peak point has thermal inertia characteristics.

[0038] Step 62: Collect continuous geomagnetic amplitude data at stable peak points to form a geomagnetic sampling sequence, and calculate the mean and standard deviation of geomagnetic amplitude based on a sliding time window. When the standard deviation is less than the preset geomagnetic steady-state threshold, it is determined that the peak point has geomagnetic steady-state fluctuation characteristics.

[0039] Step 63: Jointly verify the thermal inertia characteristic judgment result and the geomagnetic steady-state fluctuation judgment result. Only when both are satisfied is the peak point confirmed as a valid peak point, otherwise it is judged as an invalid peak point. When there are multiple similar valid peak points, the spatial Euclidean distance is less than the preset distance threshold as the similarity judgment condition, and the peak point with the largest composite difference is selected as the unique peak point, and the unique valid peak point information is output.

[0040] Furthermore, the localization result is obtained through gradient intersection of stable peak points, and other modules are called along different paths for verification, including:

[0041] Step 71: Based on the unique valid peak point information and the continuous position sequence of the handheld device, coordinate reference unification and error modeling are performed. The unique valid peak point is converted into candidate positioning coordinates in the global coordinate system, and the error radius is obtained by calculating the square root of the sampling variance of the peak point.

[0042] Step 72: Using the candidate positioning coordinates as the center, uniformly distribute the verification observation positions in their neighborhood, and perform short-range movement at each verification observation position along the direction of maximum increase of composite difference to form multiple verification direction lines; by performing least squares calculation on the verification direction lines, the intersection point is obtained as the positioning result;

[0043] Step 73: Repeat the spatiotemporal sampling acquisition module, difference correction analysis module, candidate clustering screening module, and gradient approximation confirmation module along at least two inspection paths different from the original inspection path. Re-execute step 72 within the neighborhood of the candidate positioning coordinates. If the Euclidean distance between the multiple obtained positioning results is less than the preset verification threshold, the final positioning result is confirmed; otherwise, it is marked as needing re-inspection. The different inspection paths are mutually orthogonal S-shaped paths.

[0044] The beneficial effects of this invention are as follows: By jointly acquiring and calculating the composite difference of multiple passive parameters such as temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance, this invention effectively avoids the false alarms and missed alarms caused by reliance on a single sensor; by combining correction mechanisms such as attitude angle compensation, velocity weighting, and neighborhood difference, it significantly improves the accuracy and stability of the sampled data; by introducing whitelist and shielding domain constraints, it can automatically identify and eliminate common interference sources, reducing the impact of environmental noise on the detection results; by employing multi-level processing methods such as candidate clustering, differential gradient guidance, peak domain kernel confirmation, and stable peak point screening, it ensures the uniqueness and reliability of the anomaly source location results; and by using a multi-path verification mechanism, it further verifies the consistency of the location results, improving the robustness of the system under different locations and inspection conditions. Overall, this invention can achieve high-precision, high-stability, and highly adaptable location and space safety detection. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of a handheld detection system for space safety according to the present invention. Detailed Implementation

[0046] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0047] like Figure 1 As shown, a handheld detection system for spatial safety includes:

[0048] The static fingerprint construction module is used to establish a static fingerprint database and whitelist and its shielding domain. The static fingerprint database is based on passive parameters collected according to a preset grid under a no-suspicion state.

[0049] The spatiotemporal sampling acquisition module is used to acquire continuous spatiotemporal sampling sequences through a handheld device, obtain continuous position sequences based on the handheld inspection path, and synchronously collect passive parameters at each position to form a spatiotemporal sampling sequence.

[0050] The difference correction analysis module is used to correct readings and construct composite difference for spatiotemporal sampling sequences. It is based on posture, walking speed and changes in the neighboring environment for correction. The corrected passive parameters are compared with the static fingerprint database benchmark to obtain the composite difference.

[0051] The candidate clustering and filtering module is used to perform candidate clustering and persistence filtering under whitelist constraints. It performs connectivity clustering on positions where the composite difference exceeds a preset threshold outside the shielded domain. It filters based on the drop and area conditions of the composite difference and outputs the centroid, maximum difference point and principal axis direction of the candidate cluster.

[0052] The gradient approximation confirmation module is used to perform walking approximation and peak kernel confirmation guided by differential gradient using a handheld device. It advances according to the spatial upward direction of the composite difference degree and performs circumferential mapping within the candidate cluster to confirm the peak kernel and obtain stable peak point information.

[0053] The close-range verification module is used to perform close-range passive verification and unique positioning based on thermal inertial characteristics and geomagnetic steady-state fluctuation characteristics. It obtains the positioning result through gradient intersection of stable peak points and calls other modules for verification along different paths.

[0054] In one embodiment of the present invention, the whitelist records the location and average passive parameter values ​​of common sources. Common sources refer to stable passive physical interference sources that exist long-term in the inspection area, such as fixed lighting fixtures, metal brackets, ventilation openings, or large equipment. These objects generate stable background responses during passive parameter acquisition. By recording the location coordinates and corresponding average passive parameter values ​​of such common sources into the whitelist, they can be identified and blocked during subsequent detection processes, thereby avoiding misjudgment as abnormal targets.

[0055] The shielding domain is obtained by merging the influence range of whitelisted objects; this shielding domain is used to eliminate differences caused by common sources during anomaly detection, thereby improving the accuracy and stability of detection.

[0056] The passive parameters include: temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance; visible light reflectance refers to the intensity of incident light reflected from the target location in the visible light band.

[0057] In one embodiment of the present invention, a static fingerprint database and a whitelist and its shielding domain are established, wherein the static fingerprint database collects passive parameters according to a preset grid based on a no-suspicion state, including:

[0058] Step 11: Clear the mobile electronic devices in the venue to a state of no suspicion, and divide the venue space according to the preset grid; where no suspicion means that there are no active abnormal electronic devices or suspicious interference sources in the venue, and only fixed structures and stable background signals are retained.

[0059] Step 12: Sample each grid multiple times to collect passive parameters, and calculate the mean and standard deviation of temperature, geomagnetic amplitude, electrostatic potential and visible light reflectance. Use the mean and standard deviation as the baseline statistical values ​​for that grid point and store them in the static fingerprint database. This static fingerprint database is used as a reference for calculating composite differences in subsequent detection processes.

[0060] Step 13: Record the location and corresponding baseline statistics of common sources related to passive parameters to form a whitelist, and generate corresponding shielding areas based on the influence range of each common source. Merge the shielding areas to obtain the shielding domain. By setting the whitelist and shielding domain, the interference of common sources on the detection results can be effectively avoided, thereby improving the accuracy and robustness of the detection.

[0061] In one embodiment of the present invention, a continuous spatiotemporal sampling sequence is acquired through a handheld device, a continuous position sequence is obtained according to the handheld inspection path, and passive parameters are synchronously collected at each position to form a spatiotemporal sampling sequence, including:

[0062] Step 21: The operator walks along a preset inspection path using a handheld device. This preset inspection path is an S-shaped path covering the boundary of the site and extending into the interior. A continuous position sequence is obtained by coordinating drift error control using inertial odometry and a visual loopback algorithm. Specifically, the operator carries the handheld device and walks along the preset inspection path within the site. The inspection path is designed as an S-shaped path covering the boundary of the site and extending into the interior area to ensure the integrity of spatial coverage. During the inspection, the handheld device simultaneously relies on inertial odometry and a visual loopback algorithm for position estimation. The inertial odometry is used to calculate relative displacement based on accelerometer and gyroscope signals, while the visual loopback algorithm identifies recurring scenes through environmental images captured by a camera, thereby constraining and correcting the cumulative drift generated by the inertial odometry. Through the synergistic effect of both, a continuous position sequence with controlled error can be obtained.

[0063] Step 22: At each location point in the continuous location sequence, data is collected within a preset time window. Within each time window, temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance are acquired, and the window mean is calculated to form a parameter sequence. The interference caused by instantaneous fluctuations is reduced by calculating the window mean.

[0064] Step 23: Bind the continuous position sequence to the parameter sequence one by one, and store the corresponding passive parameters using the coordinates of the position points as indexes to form a spatiotemporal sampling sequence. The spatiotemporal sampling sequence established in this way can accurately reflect the environmental passive characteristics of each position point on the inspection path, eliminate instantaneous noise in the time dimension, and maintain the precise correspondence between coordinates and parameters in the spatial dimension.

[0065] In one embodiment of the present invention, reading correction is performed on the spatiotemporal sampling sequence and a composite dissimilarity is constructed. Correction is based on posture, walking speed, and changes in the nearest neighbor environment. The corrected passive parameters are compared with a static fingerprint database benchmark to obtain the composite dissimilarity, including:

[0066] Step 31: Obtain the attitude angle of the handheld device and compensate and correct the visible light reflectance values ​​in the spatiotemporal sampling sequence. The compensation and correction uses the cosine value of the attitude angle as the correction coefficient. The attitude angle refers to the tilt angle of the device relative to the horizontal plane, measured by an accelerometer. Since the acquisition results of visible light reflectance by the device at different angles will be affected, it is necessary to compensate and correct the visible light reflectance values ​​in the spatiotemporal sampling sequence. In this embodiment, the cosine value of the attitude angle is used as the correction coefficient, which is multiplied by the original visible light reflectance value to offset the measurement deviation caused by the device tilt, thereby achieving attitude compensation of the visible light reflectance data.

[0067] Step 32: Determine the weight at each location point based on the walking speed. The weight is obtained by calculating the ratio of the walking speed at the current location to the preset reference speed. When the actual walking speed is faster or slower than the reference speed, the weight will be adjusted accordingly to reflect the stability of the sampled data in the time dimension. During the passive parameter correction process at each location point, differential correction is performed by combining the average value of the sampling points within the preset sampling radius centered on that location point to correct random interference caused by local environmental fluctuations.

[0068] Step 33: The corrected passive parameters are compared with the corresponding baseline values ​​in the static fingerprint database. The comparison uses the standard deviation of each passive parameter in the static fingerprint database as a normalization factor, and calculates the composite difference by taking the square root of the sum of the squares of the differences between each passive parameter and its corresponding baseline value. The composite difference is a comprehensive index reflecting the degree of difference between the actual sampled value and the baseline environment. The larger the value, the higher the degree to which the passive features at that location deviate from the no-suspicion state.

[0069] Through the above process, this embodiment can correct the original sampling data in multiple dimensions, including eliminating the influence of posture, compensating for velocity differences, and smoothing the environmental neighborhood, ensuring the objectivity and consistency of the comparison.

[0070] In one embodiment of the present invention, the candidate clustering screening module includes:

[0071] Step 41: Outside the shielded area, filter the locations where the composite difference exceeds a preset difference threshold to obtain a candidate point set. The preset threshold is the sum of the mean of the composite difference and the standard deviation of a preset multiple. This statistical method can dynamically adapt to environmental fluctuations and reduce oversensitivity or omissions.

[0072] Step 42: Cluster the candidate point set according to spatial connectivity to obtain a candidate cluster set. Spatial connectivity is determined based on whether the Euclidean distance between candidate points is less than a preset neighborhood radius. For each candidate cluster, calculate the centroid, maximum dissimilarity point, and principal axis direction. The principal axis direction is obtained by eigenvalue decomposition of the covariance matrix of the candidate cluster point set. Specifically, spatial connectivity refers to whether candidate points are sufficiently close in space, determined based on Euclidean distance, i.e., by calculating the straight-line distance between two candidate points in two-dimensional or three-dimensional spatial coordinates. When this distance is less than a preset neighborhood radius, the two points are considered connected. The centroid is the arithmetic mean of the coordinates of all points within the cluster, reflecting the location center of the cluster. The maximum dissimilarity point is the point with the highest composite dissimilarity value within the cluster, indicating the most significant location of the anomaly.

[0073] Step 43: Calculate the persistence of candidate clusters and filter them based on area conditions, outputting the centroid, maximum difference point, and principal axis direction of candidate clusters that meet the conditions. The persistence is the difference between the maximum composite difference within a candidate cluster and the minimum composite difference at the convex hull boundary of the candidate cluster, where the convex hull refers to the smallest convex polygon or convex polyhedron that can cover all points within the cluster. This method can measure the difference between the abnormal signal within the cluster and the boundary noise; a higher persistence indicates a more stable anomaly. The area condition is obtained by ensuring that the number of points within the candidate cluster is greater than or equal to a preset threshold.

[0074] Through the above steps, the candidate clustering screening module can effectively eliminate environmental noise and isolated interference from a large number of sampling points, retaining regions with significant spatial coherence and stable differences, thereby improving the accuracy and reliability of subsequent anomaly localization.

[0075] In one embodiment of the present invention, the differential gradient-guided walking approximation and peak region kernel confirmation includes:

[0076] Step 51: Calculate the upward direction based on the gradient of the composite difference degree within the candidate cluster, select the direction of the fastest upward movement as the travel direction, and gradually move along this direction using a handheld device to update the position points and form an approximation path sequence. The upward direction is obtained by differential calculation of the composite difference degree of adjacent position points within the candidate cluster. The gradual movement adopts an adaptive step size adjusted according to the magnitude of the composite difference degree gradient. That is, the step size is automatically adjusted according to the magnitude of the composite difference degree gradient. When the gradient is large, a smaller step size is used to improve accuracy, and when the gradient is small, a larger step size is used to improve efficiency. In this way, the local peak region can be gradually approached along the most significant difference upward direction. The gradient of the composite difference degree refers to the direction of change of the composite difference degree value in spatial distribution, which is obtained by differential calculation of the composite difference degree of adjacent position points within the candidate cluster.

[0077] Step 52: When the approximation path approaches the local peak region of the candidate cluster, sampling is performed around the peak along a closed path to form a surrounding sampling set. The composite difference of the surrounding sampling set is then interpolated on a plane to draw contour lines. The boundary of the peak region core is determined based on the shape of the contour lines. The closed path is a ring path centered on the local peak point, ensuring that the sampling points are evenly distributed around the peak. The difference processing uses linear interpolation to estimate continuous values, thereby generating continuous contour lines. The shape of the contour lines reflects the spatial distribution characteristics of the difference. The boundary of the peak region core is determined based on the rate of change of the area of ​​the closed contour region being less than a preset rate of change threshold.

[0078] Step 53: Within the peak region core, the point with the largest composite difference is identified as a candidate peak point. Sampling is repeated under different surrounding surveying paths. If the candidate peak point is detected in at least two surrounding surveying paths, and its spatial position difference is less than a preset consistency threshold, then the candidate peak point is determined as a stable peak point, and stable peak point information is output. The stable peak point information may include the spatial coordinates of the peak point, the corresponding composite difference value, and the determination results of thermal inertia characteristics and geomagnetic steady-state characteristics. The preset consistency threshold refers to the maximum allowable spatial deviation of the peak point position under different surveying paths.

[0079] Through the above design, this embodiment can effectively eliminate environmental noise interference in complex environments, achieve rapid approximation by using gradient guidance, and improve the stability and accuracy of peak point positioning by combining surrounding mapping and repeated verification, thereby providing reliable initial candidate points for subsequent close-range passive verification and unique positioning.

[0080] In one embodiment of the present invention, performing near-field passive verification includes:

[0081] Step 61: Collect continuous temperature data at the stable peak point to form a temperature sampling sequence, and calculate the rate of temperature change over time by the difference between adjacent sampling points. When the rate of temperature change is lower than the preset thermal inertia threshold, it is determined that the peak point has thermal inertia characteristics. The thermal inertia threshold is determined by a baseline temperature experiment under no suspicion conditions.

[0082] Step 62: Collect continuous geomagnetic amplitude data at stable peak points to form a geomagnetic sampling sequence, and calculate the mean and standard deviation of geomagnetic amplitude based on a sliding time window. When the standard deviation is less than a preset geomagnetic steady-state threshold, the peak point is determined to have geomagnetic steady-state fluctuation characteristics. The geomagnetic steady-state threshold is a preset multiple of the standard deviation of geomagnetic amplitude in the static fingerprint database, so that the steady-state determination range can be adaptively adjusted under different environmental conditions.

[0083] Step 63: Jointly verify the thermal inertia characteristic judgment result and the geomagnetic steady-state fluctuation judgment result. Only when both are satisfied is the peak point confirmed as a valid peak point, otherwise it is judged as an invalid peak point. When there are multiple similar valid peak points, the spatial Euclidean distance is less than the preset distance threshold as the similarity judgment condition, and the peak point with the largest composite difference is selected as the unique peak point, and the unique valid peak point information is output.

[0084] Through the above steps, this embodiment can utilize both thermal inertial characteristics and geomagnetic steady-state fluctuation characteristics to perform dual verification of peak points, thereby significantly improving the reliability of anomaly source identification. Joint verification avoids misjudgments caused by a single feature, while the unique processing of similar peak points ensures the singularity and accuracy of the final location result.

[0085] In one embodiment of the present invention, the localization result is obtained by gradient intersection of stable peak points, and other modules are called along different paths for verification, including:

[0086] Step 71: Based on the unique valid peak point information and the continuous position sequence of the handheld device, coordinate reference unification and error modeling are performed. The unique valid peak point is converted into candidate positioning coordinates in the global coordinate system, and the error radius is obtained by calculating the square root of the sampling variance of the peak point, which is used to quantify the uncertainty range of the peak point positioning. Here, coordinate reference unification refers to transforming the peak point position in the relative coordinate system to the global coordinate system, thereby ensuring that the processing results of different paths and different modules have a consistent spatial reference.

[0087] Step 72: Using the candidate positioning coordinates as the center, uniformly distribute the verification observation positions in their neighborhood, and perform short-range movements at each verification observation position along the direction of maximum increase in composite difference to form multiple verification direction lines; by performing least squares calculation on the verification direction lines, that is, by minimizing the sum of squared distances between each verification direction line and the intersection point, the intersection point is obtained as the positioning result; the direction of maximum increase in composite difference refers to the direction of the most significant increase in the composite difference gradient around the point, which is used to indicate the spatial expansion trend of the anomaly source.

[0088] Step 73: Repeat the spatiotemporal sampling acquisition module, difference correction analysis module, candidate clustering screening module, and gradient approximation confirmation module along at least two inspection paths different from the original inspection path. Re-execute step 72 within the neighborhood of the candidate positioning coordinates. If the Euclidean distance between the multiple obtained positioning results is less than the preset verification threshold, the final positioning result is confirmed; otherwise, it is marked as needing re-inspection. The different inspection paths are mutually orthogonal S-shaped paths to ensure multi-angle coverage and cross-verification of the spatial area.

[0089] Through the above design, this embodiment can effectively improve the uniqueness and reliability of positioning by combining a multi-path, multi-module, and multi-observation verification mechanism on the basis of preliminary positioning. Coordinate benchmark unification and error modeling provide quantitative constraints on the results, while the verification of observation positions and least squares calculation enhance the robustness of geometric intersection, and the multi-path verification mechanism ensures that the results do not depend on the accidental characteristics of a single path, thereby improving the overall adaptability and reliability of the detection system in complex environments.

[0090] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0091] The embodiments of the present invention have been described above, but the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of the present embodiments, all of which are within the protection scope of the present embodiments.

Claims

1. A handheld detection system for spatial safety, characterized in that, include: The static fingerprint construction module is used to establish a static fingerprint database and whitelist and its shielding domain. The static fingerprint database is based on passive parameters collected according to a preset grid under a no-suspicion state. The spatiotemporal sampling acquisition module is used to acquire continuous spatiotemporal sampling sequences through a handheld device, obtain continuous position sequences based on the handheld inspection path, and synchronously collect passive parameters at each position to form a spatiotemporal sampling sequence. The difference correction analysis module is used to correct readings and construct composite difference for spatiotemporal sampling sequences. It is based on posture, walking speed and changes in the neighboring environment for correction. The corrected passive parameters are compared with the static fingerprint database benchmark to obtain the composite difference. The candidate clustering and filtering module is used to perform candidate clustering and persistence filtering under whitelist constraints. It performs connectivity clustering on positions where the composite difference exceeds a preset threshold outside the shielded domain. It filters based on the drop and area conditions of the composite difference and outputs the centroid, maximum difference point and principal axis direction of the candidate cluster. The gradient approximation confirmation module is used to perform walking approximation and peak kernel confirmation guided by differential gradient using a handheld device. It advances according to the spatial upward direction of the composite difference degree and performs circumferential mapping within the candidate cluster to confirm the peak kernel and obtain stable peak point information. The close-range verification module is used to perform close-range passive verification and unique positioning based on thermal inertial characteristics and geomagnetic steady-state fluctuation characteristics. It obtains the positioning result through gradient intersection of stable peak points and calls other modules for verification along different paths.

2. The handheld detection system for spatial safety according to claim 1, characterized in that, The whitelist records the location of common sources and the average value of passive parameters; The shielded domain is obtained by merging the influence range of the whitelisted objects; The passive parameters include: temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance.

3. The handheld detection system for spatial safety according to claim 1, characterized in that, Establish a static fingerprint database and whitelist, along with their blocking domains. The static fingerprint database collects passive parameters according to a preset grid based on a no-suspect status, including: Step 11: Clear all mobile electronic devices from the premises until they are free of suspicion, and divide the premises space according to a preset grid. Step 12: Sample each grid multiple times to collect passive parameters, and calculate the mean and standard deviation of temperature, geomagnetic amplitude, electrostatic potential and visible light reflectance respectively. Use the mean and standard deviation as the benchmark statistical values ​​of the grid and store them in the static fingerprint database. Step 13: Record the location of common sources related to passive parameters and their corresponding baseline statistics to form a whitelist, and generate corresponding shielding areas based on the influence range of each common source. Merge the shielding areas to obtain the shielding domain.

4. The handheld detection system for spatial safety according to claim 1, characterized in that, A continuous spatiotemporal sampling sequence is acquired using a handheld device. A continuous position sequence is obtained based on the handheld inspection path, and passive parameters are simultaneously collected at each position to form the spatiotemporal sampling sequence, including: Step 21: Walk along a preset inspection path using a handheld device. The preset inspection path is an S-shaped path that covers the boundary of the site and extends into the interior. Obtain a continuous position sequence by coordinating drift error constraint based on inertial odometry and visual loop closure algorithm. Step 22: At each location point in the continuous location sequence, data is collected in a preset time window. Within each time window, temperature, geomagnetic amplitude, electrostatic potential, and visible light reflectance are acquired, and the window mean is calculated to form a parameter sequence. Step 23: Bind the continuous position sequence to the parameter sequence one by one, and store the corresponding passive parameters with the coordinates of the position points as indexes to form a spatiotemporal sampling sequence.

5. The handheld detection system for spatial safety according to claim 1, characterized in that, Reading corrections are performed on the spatiotemporal sampling sequences, and a composite dissimilarity score is constructed. Corrections are based on attitude, walking speed, and changes in the nearest neighbor environment. The corrected passive parameters are compared with a static fingerprint database benchmark to obtain the composite dissimilarity score, which includes: Step 31: Obtain the attitude angle of the handheld device and compensate and correct the visible light reflectance value in the spatiotemporal sampling sequence. The compensation and correction uses the cosine value of the attitude angle as the correction coefficient. Step 32: Determine the weight at each location point based on the walking speed. The weight is obtained by calculating the ratio of the walking speed at the current location to the preset reference speed. During the passive parameter correction process at each location point, differential correction is performed by combining the average value of the sampling points within the preset sampling radius centered on that location point. Step 33: The corrected passive parameters are compared with the corresponding baseline values ​​in the static fingerprint database. The comparison uses the standard deviation of each passive parameter in the static fingerprint database as a normalization factor, and calculates the composite difference by taking the square root of the sum of the squares of the differences between each passive parameter and the corresponding baseline value.

6. The handheld detection system for spatial safety according to claim 1, characterized in that, The candidate clustering screening module includes: Step 41: Outside the shielding domain, filter the locations where the composite difference exceeds a preset difference threshold to obtain a candidate point set. The preset difference threshold is the sum of the mean of the composite difference and the standard deviation of a preset multiple. Step 42: Cluster the candidate point set according to spatial connectivity to obtain a candidate cluster set. The spatial connectivity is determined based on whether the Euclidean distance between candidate points is less than a preset neighborhood radius. Calculate the centroid, maximum difference point, and principal axis direction for each candidate cluster. The principal axis direction is obtained by performing eigenvalue decomposition on the covariance matrix of the candidate cluster point set. Step 43: Calculate the persistence of candidate clusters and filter them in combination with the area condition, and output the centroid, maximum difference point and principal axis direction of the candidate clusters that meet the conditions; the persistence is the difference between the maximum value of the composite difference within the candidate cluster and the minimum value of the composite difference at the convex hull boundary of the candidate cluster, and the area condition is obtained by the number of points within the candidate cluster being greater than or equal to a preset threshold.

7. The handheld detection system for spatial safety according to claim 1, characterized in that, The differential gradient-guided walking approximation and peak region kernel confirmation include: Step 51: Calculate the upward direction based on the gradient of the composite difference degree within the candidate cluster, select the position direction with the fastest upward movement as the direction of travel, and gradually move along the direction of travel by carrying a handheld device to update the position points and form an approximation path sequence; the upward direction is obtained by differential calculation of the composite difference degree of adjacent position points within the candidate cluster, and the gradual movement adopts an adaptive step size adjusted according to the gradient magnitude of the composite difference degree. Step 52: When the approximation path approaches the local peak region of the candidate cluster, sampling is performed around the peak along a closed path to form a surrounding sampling set. The composite difference of the surrounding sampling set is interpolated on the plane and then contour lines are drawn. The boundary of the peak region kernel is determined according to the shape of the contour lines. The closed path is a ring path with the local peak point as the center. Step 53: Within the peak region core, determine the point with the largest composite difference as a candidate peak point. By repeatedly performing sampling under different surrounding mapping paths, if the candidate peak point is detected under at least two surrounding mapping paths and its spatial position difference is less than a preset consistency threshold, then the candidate peak point is determined as a stable peak point, and stable peak point information is output.

8. A handheld detection system for spatial safety according to claim 7, characterized in that, In step 52, the boundary of the peak region is determined based on the fact that the rate of change of the area of ​​the contour line closed region is less than a preset rate of change threshold.

9. A handheld detection system for spatial safety according to claim 1, characterized in that, Performing near-field passive verification includes: Step 61: Collect continuous temperature data at stable peak points to form a temperature sampling sequence, and calculate the rate of temperature change over time by the difference between adjacent sampling points. When the rate of temperature change is lower than a preset thermal inertia threshold, it is determined that the peak point has thermal inertia characteristics. Step 62: Collect continuous geomagnetic amplitude data at stable peak points to form a geomagnetic sampling sequence, and calculate the mean and standard deviation of geomagnetic amplitude based on a sliding time window. When the standard deviation is less than the preset geomagnetic steady-state threshold, it is determined that the peak point has geomagnetic steady-state fluctuation characteristics. Step 63: Jointly verify the thermal inertia characteristic judgment result and the geomagnetic steady-state fluctuation judgment result. Only when both are satisfied is the peak point confirmed as a valid peak point, otherwise it is judged as an invalid peak point. When there are multiple similar valid peak points, the spatial Euclidean distance is less than the preset distance threshold as the similarity judgment condition, and the peak point with the largest composite difference is selected as the unique peak point, and the unique valid peak point information is output.

10. A handheld detection system for spatial safety according to claim 1, characterized in that, The localization result is obtained by gradient intersection of stable peak points, and other modules are called along different paths for verification, including: Step 71: Based on the unique valid peak point information and the continuous position sequence of the handheld device, coordinate reference unification and error modeling are performed. The unique valid peak point is converted into candidate positioning coordinates in the global coordinate system, and the error radius is obtained by calculating the square root of the sampling variance of the peak point. Step 72: Using the candidate positioning coordinates as the center, uniformly distribute the verification observation positions in their neighborhood, and perform short-range movement at each verification observation position along the direction of maximum increase of composite difference to form multiple verification direction lines; by performing least squares calculation on the verification direction lines, the intersection point is obtained as the positioning result; Step 73: Repeat the spatiotemporal sampling acquisition module, difference correction analysis module, candidate clustering screening module, and gradient approximation confirmation module along at least two inspection paths different from the original inspection path. Re-execute step 72 within the neighborhood of the candidate positioning coordinates. If the Euclidean distance between the multiple obtained positioning results is less than the preset verification threshold, the final positioning result is confirmed; otherwise, it is marked as needing re-inspection. Among them, the inspection paths different from the original inspection path are mutually orthogonal S-shaped paths.

Images