Electromagnetic field near-field visualization method and device fusing machine vision and electromagnetic measurement

Abstract

The application discloses an electromagnetic field near-field visualization method and device fusing machine vision and electromagnetic measurement, and relates to the technical field of electromagnetic compatibility test and electromagnetic environment perception. The method firstly acquires an image of a surface of a measured object, identifies and positions a probe by using a model, and divides a field of view area into grids; meanwhile, electromagnetic field intensity data are acquired, and representative typical field intensity values are extracted. When the probe is identified as being located in a certain grid unit, if the grid is entered for the first time, the grid is created and a typical value of a current area is written; if the grid already exists, the typical value of the grid is updated on the basis of the existing grid, and a two-dimensional and three-dimensional electromagnetic field diagram in continuous distribution is formed. The interpolation result of the electromagnetic field is fused and rendered with a video image, and the visualization output of electromagnetic field distribution in a real scene is realized. The method can ensure the measurement accuracy, and intuitively presents the electromagnetic field distribution, the three-dimensional structure of equipment and the actual environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electromagnetic compatibility testing and electromagnetic environment perception, and particularly relates to an electromagnetic field near-field visualization method fusing machine vision and electromagnetic measurement. BACKGROUND

[0002] With the development of high-speed, miniaturization and high integration of electronic devices, the working frequency of circuits is continuously increasing, the device spacing is continuously reducing, and the electromagnetic coupling effect of the system is dramatically enhanced. Electromagnetic compatibility (EMC) problems show a multi-source and complex trend, which brings severe challenges to the stable operation and electromagnetic protection of electronic systems. Various high-frequency and high-speed circuits, wireless communication modules and power electronic devices will produce electromagnetic interference with different intensities and wide frequency bands during operation. If not effectively controlled, it may interfere with surrounding sensitive devices and systems, and even threaten the safe operation of the equipment and system.

[0003] In order to accurately evaluate the electromagnetic leakage intensity and potential interference source distribution of electronic devices, electromagnetic field near-field scanning technology has become an indispensable important means in electromagnetic compatibility testing and design optimization. By scanning the near-field region on the surface of the measured object, the amplitude, phase and direction distribution information of the electromagnetic field can be obtained, so as to realize the accurate positioning of the electromagnetic interference source and the coupling path, and provide quantitative basis for subsequent structure optimization and interference suppression design.

[0004] However, although the traditional near-field scanning technology can obtain local electromagnetic field data, the results are mostly presented in the form of amplitude-frequency characteristics or two-dimensional discrete graphs, and lack of association with the three-dimensional structure of the measured object and the actual spatial environment. The test data stays at the abstract numerical level, and the engineers need to repeatedly compare between the electromagnetic distribution graph and the actual device when analyzing, which is difficult to intuitively judge the spatial position and cause of electromagnetic anomalies. In addition, the existing visualization means are mostly completed in the post-processing stage, and cannot realize the synchronous fusion of measurement data and actual scene information, making it difficult to reveal the corresponding relationship between electromagnetic field characteristics and structure morphology, and limiting the application value of test results in design optimization and electromagnetic protection analysis. SUMMARY

[0005] The technical problem to be solved by the present application is to provide an electromagnetic field near-field visualization method fusing machine vision and electromagnetic measurement, which can intuitively present the electromagnetic field distribution, the three-dimensional structure of the device and the actual environment while ensuring measurement accuracy.

[0006] To solve the above technical problems, the technical solution adopted by the present application is: an electromagnetic field near-field visualization method fusing machine vision and electromagnetic measurement, comprising the following steps:

[0007] The image acquisition device continuously acquires the image of the surface of the measured object, the target detection model is used for identifying and positioning the probe, and the field region is divided into grid units;

[0008] The API of the electromagnetic field strength measurement device is called to acquire the electromagnetic field strength data of the measured object in real time, and the historical data is accumulated to extract representative field strength typical values;

[0009] When the probe is identified as being located in a certain grid unit, if the grid unit is entered for the first time, the grid is created and the field strength typical value of the current region is written; if it already exists, the field strength typical value of the grid is updated based on the existing grid;

[0010] After the field strength data of the grid unit is updated, the laser ranging module is called to acquire the distance information between the probe and the measured object, and the field strength is normalized according to the distance compensation model;

[0011] After the data collection of each grid unit is completed, the inverse distance weighted interpolation algorithm is used to combine the smoothing constraint model to reconstruct the spatial field strength, and the continuous distribution of two-dimensional and three-dimensional electromagnetic field maps is formed;

[0012] The electromagnetic field interpolation result is fused and rendered with the video image to realize the visualization output of the electromagnetic field distribution in the real scene.

[0013] The application further discloses an electromagnetic field near-field visualization device fusing machine vision and electromagnetic measurement, the device uses the electromagnetic field near-field visualization method, and the device comprises:

[0014] The probe identification unit is used for acquiring the surface image of the measured object through the camera and positioning the near-field probe by using the target detection algorithm;

[0015] The grid generation and data bearing unit is used for dynamically generating regularly arranged grid regions according to the probe identification result, each grid region is used as a data container for storing, managing and bearing the subsequent collected electromagnetic field test data, and the adaptive update of spatial division is realized;

[0016] The data acquisition and bearing unit is used for calling the Python interface to access the API of the electromagnetic field measuring instrument, continuously receiving the magnetic field amplitude data of the corresponding position, and storing the data in the data container of the grid unit in real time;

[0017] The statistical modeling and feature value establishing unit is used for performing distribution analysis on the discrete magnetic field data set recorded in each grid unit, and extracting a representative data center value based on a normal model;

[0018] The probe distance measuring unit is used for acquiring the instantaneous distance between the probe and the surface of the measured object in real time through the integrated laser range finder, and binding and storing the distance with the corresponding magnetic field amplitude data one by one.

[0019] The amplitude normalization compensation unit is used for performing amplitude normalization operation on each piece of collected data based on the frequency-dependent power law decay model obtained through calibration, so as to make it equivalent to be mapped to a unified reference height.

[0020] The spatial interpolation modeling unit is used for taking the dB amplitude in each grid region as input, and performing continuous field reconstruction according to the spatial topological relationship.

[0021] The visual fusion rendering unit is used for mapping the interpolated electromagnetic distribution matrix to the surface image obtained by the real camera, or superimposing it in the three-dimensional grid model of the measured object, to complete the near-field visualization of the electromagnetic field.

[0022] The beneficial effects produced by the above technical solutions are that the method described in the application can realize the near-field scanning visualization technology of the fusion of electromagnetic data and spatial visual information, can intuitively present the electromagnetic field distribution and the three-dimensional structure of the equipment and the actual environment while ensuring the measurement accuracy, and provides reliable and easy-to-understand technical support for electromagnetic compatibility testing, electromagnetic interference tracing and engineering design optimization. BRIEF DESCRIPTION OF DRAWINGS

[0023] The application will be described in further detail below with reference to the drawings and specific embodiments.

[0024] Figure 1 is a flowchart of the method described in the embodiments of the application;

[0025] Figure 2 is a principle diagram of laser ranging in the method described in the embodiments of the application;

[0026] Figure 3 is a scanning effect diagram in the method described in the embodiments of the application;

[0027] Figure 4 is a module composition structure diagram of the device described in the embodiments of the application. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0029] In the following description, a lot of specific details are set forth in order to provide a thorough understanding of the present application, however, the present application can be practiced in other manners different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the present application, therefore, the present application is not limited to the specific embodiments disclosed below.

[0030] Embodiment one

[0031] As shown in the figure, the embodiment of the present application discloses an electromagnetic field near-field visualization method combining machine vision and electromagnetic measurement, including the following steps: Figure 1 The image acquisition device continuously acquires the image of the surface of the measured object, the target detection model is used to identify and locate the probe, and the field of view area is divided into grid units;

[0032] The API of the electromagnetic field intensity measuring device is called to acquire the electromagnetic field intensity data of the measured object in real time, and the historical data is accumulated to extract representative field intensity typical values;

[0033] When the probe is identified as being located in a certain grid unit, if the grid unit is entered for the first time, the grid is created and the field intensity typical value of the current area is written; if it already exists, the field intensity typical value of the grid is updated based on the existing grid;

[0034] After the field intensity data of the grid unit is updated, the laser ranging module is called to acquire the distance information between the probe and the measured object, and the field intensity is normalized according to the distance compensation model;

[0035] After the data collection of each grid unit is completed, the inverse distance weighted interpolation algorithm is used to combine the smooth constraint model to reconstruct the spatial field intensity, forming a continuous distribution of two-dimensional and three-dimensional electromagnetic field diagram;

[0036] The electromagnetic field interpolation result is fused and rendered with the video image to realize the visualization output of the electromagnetic field distribution in the real scene.

[0037] The above steps will be described in detail in combination with specific contents:

[0038] Near-field scanning surface gridding area division based on visual recognition:

[0039] The present application provides a near-field scanning surface gridding area division method based on visual recognition, which is used in an electromagnetic field near-field scanning device, realizes automatic grid division of the surface of the measured object, and provides a unified data container for subsequent test data, mainly including probe identification, grid generation and data bearing two parts.

[0040]

[0041] ​Probe recognition: The surface image of the measured object is collected by the camera, and the improved YOLOv8 target detection algorithm is used for near-field probe positioning. For small size targets and complex scenes, this algorithm introduces feature adaptive extraction and center region constraint strategy to improve the accuracy and robustness of the spatial position of the probe.

[0042] Specifically, for probe recognition, the application introduces a lightweight dynamic convolution module in the YOLOv8 backbone network to adaptively extract features of different scales, and embeds a channel and spatial joint attention mechanism in the feature fusion stage, thereby enhancing the recognition ability of the probe in complex backgrounds. At the same time, in the detection output stage, only the center region of the bounding box is taken as the spatial representation point to reduce the influence of edge noise and occlusion interference on probe positioning, so that the probe position remains high precision and stability in a dynamic scanning environment.

[0043] Grid generation and data carrying: According to the probe recognition results, regularly arranged grid regions are dynamically generated, and each grid region is used as a data container to store, manage and carry the subsequent collected electromagnetic field test data, realizing adaptive update of spatial division.

[0044] Specifically, for grid generation and data carrying, the application continuously collects probe center coordinates in video frames and updates grid regions in real time according to position changes, realizing dynamic spatial unit division of the entire scanning surface. Each grid unit corresponds to a unique probe sampling position, providing a stable foundation for subsequent electromagnetic field data filling, management and visualization. This mechanism not only automatically maps the spatial position of the probe to the scanning surface, but also suppresses errors caused by unstable probe movement or detection boundary drift, improving the continuity and reliability of grid division while maintaining system real-time performance to meet the requirements of rapid scanning.

[0045] Through the grid region division of the near-field scanning surface based on visual recognition of the application, the following can be achieved:

[0046] (1) Rapid and automatic positioning of near-field probes on the surface of the measured object without manual intervention;

[0047] (2) The regularly generated grid regions provide a unified data container for subsequent electromagnetic field test data, realizing systematic storage and management of data and providing a foundation for rapid scanning, analysis and visualization;

[0048] (3) Preliminary fusion of electromagnetic field distribution information and actual visual scene improves the intuitiveness of test results and engineering application value, significantly improving the efficiency and operational convenience of near-field scanning.

[0049] Electromagnetic test data statistics and typical value extraction:

[0050] After completing the gridding space division, the application proposes electromagnetic test data statistics and typical value extraction, which is used for inductive and statistical analysis of the electromagnetic amplitude data collected multiple times in the same grid area, so as to obtain the representative regional electromagnetic typical value with noise resistance. This step includes two parts of data collection bearing and statistical analysis and feature value extraction, the former is used for centralized storage and management of the data collected multiple times, and the latter extracts the typical value capable of representing the electromagnetic characteristics of the region by processing the data through a statistical method.

[0051] Data collection bearing: by calling the Python interface to access the Hidroquimica spectrum analyzer API, the magnetic field amplitude data corresponding to the position is continuously received during the probe resides in a certain one-ninth center frame area, and is stored in the data container of the grid unit in real time. Since the residence time of the probe in different areas is not fixed, the sampling data in the same area is distributed in quantity and amplitude. This unit adopts a full-cumulative recording strategy, does not directly judge or cut off the instantaneous data, and reserves the complete data basis for subsequent modeling.

[0052] Statistical modeling and feature value establishment: distribution analysis is performed on the discrete magnetic field data set recorded in each grid unit, and a representative data center value is extracted based on the normal model:

[0053] Suppose that m magnetic field amplitude data are collected in the nth spatial area, denoted as {x1,x2,…,x m}. First, the frequency statistics of the set is performed to construct the empirical distribution function f(x), and then the probability density function fitting is performed on the frequency distribution curve to obtain the approximate probability distribution function p(x), which can be expressed as:

[0054] (1)

[0055] Where p(x) is the probability density function of the field strength distribution; m is the total number of collected magnetic field amplitude data; x i represents the electromagnetic field strength observation value collected for the th time; represents the Dirac function.

[0056] By calculating the cumulative probability of the distribution, the cumulative distribution function can be obtained:

[0057] (2)

[0058] Where is the cumulative distribution function used to describe the probability that the magnetic field strength is less than or equal to a certain threshold ; is the probability density function of the field strength distribution.

[0059] The present application takes the quantile value x of the cumulative probability of 90% 0.9 As a typical representative value of the region, the data container for filling the current grid region. This selection method effectively avoids the influence of low amplitude noise, while suppressing the deviation caused by accidental peaks, so that each spatial region has a stable statistical significance field strength marker. That is:

[0060] (3)

[0061] Wherein is the typical field strength representative value of the region; is the 90% quantile function.

[0062] Through the electromagnetic test data statistical modeling and dynamic typical value extraction module of the present application, the following can be achieved:

[0063] (1) Standardized expression of electromagnetic amplitude in space, so that each grid element has clear and reproducible representative data, providing a consistent numerical benchmark for subsequent distance compensation, interpolation modeling and visualization rendering.

[0064] (2) Dynamic coverage update mechanism, when the probe passes through the same region again in the subsequent scanning process, the modeling process will be executed again with the newly collected data and the old value will be overwritten, avoiding the solidification of historical errors caused by early noise, device drift or environmental changes, so that the overall modeling system has long-term self-optimization capability.

[0065] Distance compensation and amplitude normalization calibration:

[0066] After completing the statistical establishment of electromagnetic typical values, the present application further eliminates the amplitude deviation caused by height fluctuation of the probe during scanning through distance compensation and amplitude normalization calibration, so that the measurement results at different times and different positions are established on a unified spatial scale. This step is completed through probe distance measurement and amplitude normalization compensation. The former is responsible for recording the instantaneous distance between the probe and the measured surface in real time, and the latter performs equivalent conversion and normalization processing on the original amplitude based on the distance information, thereby realizing consistent processing of data under height change conditions.

[0067] Probe distance measurement: as shown in Figure 2 , the instantaneous distance between the probe and the measured object surface is obtained in real time by integrating the laser range finder, and the distance and the corresponding magnetic field amplitude data are stored one by one. Figure 2 D1 and D2 in the formula represent the distance between the laser range finder at different positions and the measured object. Due to the high geometric sensitivity of near-field electromagnetic coupling, even millimeter-level displacement can significantly change the received amplitude, so this unit adopts a millisecond-level sampling strategy to ensure that the height recording and amplitude acquisition have strict time sequence consistency.

[0068] Amplitude normalization compensation: based on the frequency-dependent power-law decay model obtained by calibration, the amplitude normalization operation is performed on each piece of collected data, so that it is equivalent to mapping to a unified reference height. According to classical electromagnetic theory, in the quasi-static region dominated by magnetic dipoles, the magnetic field intensity and the detection distance approximately satisfy the power-law decay relationship , where ideally , that is, the field strength at 1 cm is about 8 times that at 2 cm. However, in actual engineering scenarios, the measured target is often not a standard point dipole, but a complex structure composed of PCB boards, cables, motors, shielding shells, etc., and its near-field electromagnetic distribution characteristics are affected by multiple factors such as frequency, coupling path, reflected echo, etc., so that the power index presents frequency dependence, and may be between 1 and 3. This step does not simply use a fixed for compensation, but introduces a frequency response height decay model based on calibration.

[0069] Specifically, the instantaneous distance between the probe and the target surface is obtained in real time by a laser range finder , and the actually measured amplitude is normalized to the reference height , and its compensation model is represented as:

[0070] (4)

[0071] where is the normalized field strength value; is the measured field strength value; h is the vertical height of the probe from the measured object surface; f is the frequency; is the ratio of the measured height to the normalized reference height; is the propagation attenuation index.

[0072] The present application finally uses dB as a unit to represent the field strength, and the compensation form corresponds to:

[0073] (5)

[0074] where is the normalized field strength value in dB; is the measured field strength value in dB; is the height compensation term; f is the frequency; is the ratio of the measured height to the normalized reference height.

[0075] The compensation operation is completed before data is written into the grid unit, ensuring that the subsequent statistical modeling and typical value extraction stage only processes data that has been normalized to a unified height reference. When the probe passes through the same area again, this module will re-execute the same normalization compensation process on the new data to achieve long-term self-correcting ability.

[0076] Through the distance compensation and amplitude normalization calibration module of the application, the following can be achieved:

[0077] (1) Systematic amplitude deviation introduced by height fluctuation during probe scanning is eliminated, making the measurement results at different positions and different times have a unified reference scale;

[0078] (2) Each piece of electromagnetic test data completes amplitude normalization processing before being written into the grid area, avoiding mixed dimensions or amplitude drift problems in subsequent statistical modeling;

[0079] (3) The normalization compensation strategy has adaptive updating ability. When the probe repeatedly passes through the same area, the compensation parameters can be automatically adjusted based on new data, thereby maintaining consistency and reproducibility of modeling in the long term;

[0080] (4) Provides a stable physical layer foundation for subsequent interpolation modeling and visualization rendering, making the data in each area have good lateral comparison and longitudinal traceability.

[0081] Interpolation reconstruction and scene visualization:

[0082] After completing the grid typical value establishment and distance compensation, the application constructs a continuous spatial distribution model based on the normalized electromagnetic amplitude data through the interpolation reconstruction and scene visualization module, and performs fusion rendering with the real visual scene to realize the intuitive presentation of the electromagnetic near field. Including spatial interpolation modeling and visual fusion rendering, the former is responsible for constructing a continuous electromagnetic distribution field according to the representative values in the discrete grid, and the latter maps the interpolation results to the real camera screen or three-dimensional model surface to realize the synchronous display of electromagnetic intensity and physical structure.

[0083] Spatial interpolation modeling: Take the dB amplitude in each grid area as input, and reconstruct the continuous field according to the spatial topological relationship. Considering the smoothness and direction consistency of the near-field magnetic field in the local area, the application selects the inverse distance weighting (IDW) model based on weighted distance to interpolate the representative values in each area. Its basic form is:

[0084] (6)

[0085] Where, f i is the typical amplitude of the i-th grid unit, d iwhere p is the attenuation exponent, and d is the distance between the predicted position and the center of the cell.

[0086] To further improve continuity, the application introduces a local fitting smoothing factor to constrain the second-order gradient of the interpolation result, so that the predicted surface remains derivable at the grid boundary, avoiding obvious faults or mutations.

[0087] Visual fusion rendering: map the interpolated electromagnetic distribution matrix to the surface image obtained by the real camera, or superimpose it on the three-dimensional grid model of the measured object. Specifically, the application encodes different amplitude intervals using color gradient, maps field strength regions below the set threshold to cold colors, and maps field strength regions above the set threshold to warm colors. Through the Alpha channel, semi-transparent superimposition is achieved, allowing users to observe the actual shape structure while perceiving electromagnetic intensity differences. In addition, the system supports multi-frame superimposed display at key frequency points, that is, after generating independent heat maps for different frequency bands, it plays them in time sequence to show the distribution rule of field strength with frequency change.

[0088] In the test verification stage, the application selects a single-frequency antenna composed of a loop coil as the measured object, and performs near-field scanning on it through signal excitation, obtaining the scanning effect as shown in Figure 3 .

[0089] Through the interpolation reconstruction and scene visualization module of the application, the following can be achieved:

[0090] (1) Discrete grid data is converted into a continuous near-field distribution model, making the electromagnetic field strength present a spatial gradient and diffusion pattern with physical meaning;

[0091] (2) The interpolation result is fused and displayed with the real visual scene, making the electromagnetic distribution no longer stay on the abstract numerical level, but have structure correspondence and spatial intuitiveness;

[0092] (3) Provide visual decision basis for subsequent optimization layout, shielding rectification or interference positioning, significantly improving engineering application efficiency and interactive experience.

[0093] Embodiment Two

[0094] As shown in Figure 4 , the application embodiment discloses an electromagnetic field near-field visualization device integrating machine vision and electromagnetic measurement, which comprises:

[0095] Probe recognition unit 101: used for acquiring the surface image of the measured object through the camera, and positioning the near-field probe using the improved YOLOv8 target detection algorithm;

[0096] Grid generation and data bearing unit 102: for dynamically generating regularly arranged grid areas according to the probe identification result, each grid area as a data container for storing, managing and bearing the subsequent collected electromagnetic field test data, realizing adaptive update of spatial division.

[0097] Data acquisition bearing unit 103: for accessing the Haideluo spectrum analyzer API by calling the Python interface, continuously receiving the magnetic field amplitude data corresponding to the position during the probe resides in a certain one-ninth center frame area, and storing it in the data container of the grid unit in real time.

[0098] Statistical modeling and eigenvalue establishment unit 104: for performing distribution analysis on the discrete magnetic field data set recorded in each grid unit, and extracting the representative data center value based on the normal model.

[0099] Probe distance measurement unit 105: for obtaining the instantaneous distance between the probe and the surface of the measured object in real time through the integrated laser range finder, and storing the distance and the corresponding magnetic field amplitude data one by one.

[0100] Amplitude normalization compensation unit 106: for calibrating the obtained frequency-dependent power law decay model, and performing amplitude normalization operation on each piece of collected data to make it equivalent to the unified reference height.

[0101] Spatial interpolation modeling unit 107: for taking the dB amplitude in each grid area as input, and performing continuous field reconstruction according to the spatial topological relationship.

[0102] Visual fusion rendering unit 108: for mapping the interpolated electromagnetic distribution matrix to the surface image obtained by the real camera, or superimposing it on the three-dimensional grid model of the measured object.

[0103] It should be noted that the specific implementation method of each unit in the device can refer to the specific implementation steps of the method described in Embodiment 1, which will not be repeated here.

[0104] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant regulations.

[0105] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to a memory, a database or other medium used in the embodiments provided in the present application can include at least one of a non-volatile and a volatile memory. The non-volatile memory can include a read-only memory (ROM), a magnetic tape, a floppy disk, a flash memory, an optical storage, a high-density embedded non-volatile memory, a resistive memory (ReRAM), a magnetoresistive random access memory (MRAM), a ferroelectric memory (FRAM), a phase change memory (PCM), a graphene memory, etc. The volatile memory can include a random access memory (RAM) or an external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as a static random access memory (SRAM) or a dynamic random access memory (DRAM), etc.

[0106] The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general processor, a central processing unit, a graphics processing unit, a digital signal processor, a data processing logic of a programmable logic device, etc., without being limited thereto.

[0107] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, however, as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0108] The principles and implementation modes of the present application are described by applying specific examples in the present application. The above embodiments are only used to help understand the method and its core idea of the present application; meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation mode and application range can be changed. In conclusion, the content of the present application should not be understood as a limitation.

Claims

1. A method of electromagnetic field near-field visualization merging machine vision and electromagnetic measurements, characterized in that The method comprises the following steps: continuously acquiring images of the surface of the measured object by an image acquisition device, recognizing and positioning the probe by using a target detection model, and dividing the field of view region into grid cells; calling an API of an electromagnetic field strength measurement device to acquire electromagnetic field strength data of the measured object in real time, and accumulating historical data to extract representative field strength typical values; when the probe is recognized as being located in a certain grid cell, if the grid cell is entered for the first time, the grid is created and the field strength typical values of the current region are written; if the grid cell already exists, the field strength typical values of the grid are updated based on the existing grid; after the field strength data of the grid cell are updated, a laser ranging module is called to acquire distance information between the probe and the measured object, and the field strength is normalized according to a distance compensation model; after the data collection of each grid cell is completed, a reverse distance weighted interpolation algorithm is used to reconstruct the spatial field strength in combination with a smoothing constraint model to form a continuous distribution of two-dimensional and three-dimensional electromagnetic field maps; the electromagnetic field interpolation results are fused and rendered with video images to realize the visualization output of the electromagnetic field distribution in the real scene.

2. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements according to claim 1, wherein, The method for recognizing and positioning the probe comprises the following steps: a lightweight dynamic convolution module is introduced into a YOLOv8 backbone network, a channel and spatial joint attention mechanism is embedded in a feature fusion stage, and only the center region of a bounding box is taken as a spatial representation point in a detection output stage.

3. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as claimed in claim 1, wherein, The method for dividing the field of view region into grid cells comprises the following steps: regularly arranged grid regions are dynamically generated according to the probe recognition results, each grid region is used as a data container to store, manage and carry subsequent electromagnetic field test data, and adaptive update of spatial division is realized.

4. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as claimed in claim 1, wherein, The method for acquiring electromagnetic field strength data of the measured object in real time comprises the following steps: by calling a Python interface to access an API of an electromagnetic field strength measurement device, magnetic field amplitude data corresponding to a certain center frame region where the probe resides are continuously received during the period when the probe resides in the center frame region, and the data are stored in the data container of the grid cell in real time.

5. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as claimed in claim 4, wherein, The electromagnetic field strength measurement device uses a spectrum analyzer.

6. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as claimed in claim 1, wherein, distribution analysis is performed on the discrete magnetic field data set recorded in each grid cell, and a representative data center value is extracted based on a normal model: In the n th space region, m magnetic field amplitude data are collected, denoted as {x1, x2, …, x m}First, the frequency statistics of the set are performed to construct the empirical distribution function f(x). Subsequently, the probability density function fitting is performed on the frequency distribution curve to obtain the approximate probability distribution function p(x), which is expressed as: (1) where p(x) is the probability density function of the field strength distribution; m is the total number of the magnetic field amplitude data collected so far; x i represents the electromagnetic field strength observation value collected for the first time; represents the electromagnetic field strength observation value collected for the nth time; represents the Dirac function; cumulative distribution functions are obtained by calculating the cumulative probability of the distribution: (2) wherein is a cumulative distribution function for describing the probability that the magnetic field strength takes on a value less than or equal to some threshold ; is the probability density function of the field strength distribution. The quantile value x at the cumulative probability of 90% 0.9 As a typical field strength representative value of this area, the data container for filling the current grid area makes each spatial area have a stable statistically meaningful field strength label : (3) wherein is a typical field strength representative value for the area; is a 90% quantile value function.

7. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as defined in claim 1, wherein, The method for normalizing the field strength according to the distance compensation model comprises the following steps: The instantaneous distance between the probe and the target surface is obtained in real time by a laser range finder and the actual measured amplitude is normalized to the reference height which compensates the model is represented as: (4) wherein is the normalized field strength value; is the measured field strength value; h is the vertical height of the probe from the surface of the object being measured; f is the frequency; is the ratio of the measured height to the normalized reference height; is the propagation attenuation exponent; the field strength is expressed in dB units, and the compensation form is as follows: (5) wherein is the normalized field strength value in dB; is the measured field strength value in dB; is the height compensation term; f is the frequency; is the ratio of the measured height to the normalized reference height; The compensation operation is completed before the data are written into the grid cell, and when the probe passes through the same region again, the same normalization compensation process is performed on the newly added data.

8. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as defined in claim 1, wherein, The reverse distance weighted interpolation algorithm comprises the following steps: dB amplitudes in each grid region are taken as input, the continuous field reconstruction is performed on the amplitudes according to the spatial topological relationship, a reverse distance interpolation model based on weighted distance is used to interpolate the representative values of each region, and the basic form is as follows: (6) wherein f i is the field strength typical value of the i-th grid cell, d i is the distance between the position to be predicted and the center of the cell, and p is the attenuation exponent. a local fitting smoothing factor is introduced to constrain the second-order gradient of the interpolation results, so that the predicted surface is derivable at the grid boundary.

9. The electromagnetic field near-field visualization method that fuses machine vision with electromagnetic measurements as defined in claim 1, wherein, The method for fusing and rendering the electromagnetic field interpolation results with video images comprises the following steps: The interpolated electromagnetic distribution matrix is mapped to the surface image obtained by the real camera or superimposed on the three-dimensional mesh model of the measured object, different amplitude intervals are encoded in a color gradient manner, field strength regions below a set threshold are mapped to cold colors, field strength regions above the set threshold are mapped to warm colors, and semi-transparent superimposition is realized through an Alpha channel, so that the user can perceive the electromagnetic intensity difference while observing the actual shape structure, and multi-frame superimposed display at key frequency points is supported, independent heat maps are generated for different frequency bands, and then played in a time sequence to show the distribution rule of the field strength with the change of the frequency.

10. An electromagnetic field near-field visualization apparatus that fuses machine vision with electromagnetic measurements, said apparatus using the electromagnetic field near-field visualization method as claimed in any one of claims 1-9, characterized in that The device comprises: a probe identification unit for acquiring a surface image of the measured object by a camera and positioning a near-field probe by using a target detection algorithm; a mesh generation and data bearing unit for dynamically generating regularly arranged mesh regions according to the probe identification result, each mesh region serving as a data container for storing, managing and bearing the subsequently collected electromagnetic field test data, and realizing adaptive update of spatial division; a data acquisition and bearing unit for continuously receiving the magnetic field amplitude data of the corresponding position by calling a Python interface to access the API of the electromagnetic field measuring instrument, and storing the data in the data container of the mesh unit in real time; a statistical modeling and characteristic value establishing unit for performing distribution analysis on the discrete magnetic field data set recorded in each mesh unit, and extracting a representative data center value based on a normal model; a probe distance measurement unit for acquiring the instantaneous distance between the probe and the surface of the measured object in real time by an integrated laser range finder, and storing the distance and the corresponding magnetic field amplitude data one by one; an amplitude normalization compensation unit for performing amplitude normalization operation on each piece of collected data based on the frequency-dependent power law decay model obtained through calibration, so as to make it equivalent to be mapped to a unified reference height; a spatial interpolation modeling unit for taking the dB amplitude in each mesh region as input and performing continuous field reconstruction according to the spatial topological relationship; a visual fusion rendering unit for mapping the interpolated electromagnetic distribution matrix to the surface image obtained by the real camera or superimposing it in the three-dimensional mesh model of the measured object, to complete the near-field visualization of the electromagnetic field.

Images