A multi-modal electromagnetic field distribution mapping drone system and its application method

Abstract

The present invention relates to the field of drone application technology and discloses a multimodal electromagnetic field distribution mapping drone system and its application method. The system includes a dynamic adjustment module, a multi-band acquisition module, a hybrid positioning module, a three-dimensional modeling module, and a multimodal fusion module. The dynamic adjustment module adjusts the distance between the electromagnetic sensor and the rotor according to the rotor electromagnetic interference suppression rule; the multi-band acquisition module collects and processes electromagnetic signals; the hybrid positioning module fuses TDOA and AOA algorithms to locate the radiation source; the three-dimensional modeling module generates a three-dimensional electromagnetic field intensity distribution map; and the multimodal fusion module fuses visible light images with electromagnetic data. The system can also assess radiation source risks and plan flight paths. The application method includes the steps of adjusting the distance, collecting and processing data, locating the radiation source, and constructing a model. The present invention effectively suppresses rotor interference, improves positioning accuracy, and realizes multimodal data fusion. It has broad application prospects in fields such as power inspection and urban electromagnetic environment assessment.

Description

Technical Field

[0001] The present invention relates to the field of drone application technology, and in particular to a multi-modal electromagnetic field distribution mapping drone system and an application method thereof. Background Art

[0002] With the rapid development of modern science and technology, electromagnetic fields are increasingly used in various fields, such as communications, electricity, and industrial production. However, the widespread presence of electromagnetic fields also brings many problems, making accurate mapping and analysis of their distribution crucial.

[0003] Traditional electromagnetic field mapping relied primarily on manual field measurements. Workers carried bulky measuring equipment and measured electromagnetic field strength point by point in complex environments. This method was not only inefficient but also posed significant safety risks in dangerous or difficult-to-reach areas, such as near high-voltage power towers and in mountainous areas. For example, manual measurements near high-voltage transmission lines were susceptible to interference from strong electromagnetic fields, posing a potential threat to the health of surveyors and making it difficult to guarantee the accuracy of the measured data.

[0004] Later, some measurement methods based on fixed monitoring stations emerged. While these stations can continuously monitor electromagnetic field strength within a certain range, their coverage is limited and cannot fully cover complex and changing areas. Furthermore, fixed monitoring stations are expensive to build and maintain, and it is difficult to flexibly adjust their locations to meet actual needs. For example, during urban construction, new electromagnetic radiation sources continue to emerge, and fixed monitoring stations are unable to effectively monitor these new sources in a timely manner.

[0005] In recent years, with the continuous maturity of drone technology, drones have gradually gained application in electromagnetic field mapping. However, existing drone electromagnetic field mapping systems still have many shortcomings. First, the rotation of drone rotors generates electromagnetic interference, which affects the measurement accuracy of electromagnetic sensors. While some current systems recognize this problem, the anti-interference measures they implement are ineffective and cannot fundamentally eliminate the impact of rotor electromagnetic interference on measurement data. Furthermore, existing positioning algorithms lack high accuracy for locating radiation sources. Some systems use only a single positioning algorithm, such as those based on time difference of arrival (TDOA) or angle of arrival (AOA), which are susceptible to interference from environmental factors, resulting in large positioning errors. Furthermore, most existing drone electromagnetic field mapping systems focus solely on electromagnetic data collection and analysis and lack the ability to integrate multimodal data. This inability to effectively combine electromagnetic data with other information, such as visible light imagery, makes it difficult to comprehensively analyze electromagnetic field distribution from multiple perspectives, limiting the in-depth understanding and application of electromagnetic environments.

[0006] In practical applications, such as power facility inspections, precise understanding of electromagnetic field distribution is required to detect equipment anomalies such as leakage. Urban electromagnetic environment assessments require a comprehensive consideration of geographic information and electromagnetic data to plan electromagnetically sensitive areas. Existing technologies are unable to meet these complex requirements. Therefore, the development of a multimodal electromagnetic field distribution mapping drone system and its application method is urgent, capable of effectively suppressing rotor electromagnetic interference, improving the accuracy of radiation source positioning, and enabling multimodal data fusion analysis. Summary of the Invention

[0007] The purpose of the present invention is to provide a multimodal electromagnetic field distribution mapping UAV system and its application method to solve the problems raised in the above background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: a multi-modal electromagnetic field distribution mapping UAV system, the system comprising:

[0009] The dynamic adjustment module is used to adjust the distance between the electromagnetic sensor and the UAV rotor through the retractable carbon fiber bracket according to the preset rotor electromagnetic interference suppression rules and generate dynamic distance adjustment instructions;

[0010] The multi-band acquisition module is used to control the high-precision three-axis electromagnetic field sensor to collect electromagnetic field signals within a preset frequency band, and perform anti-interference filtering on the collected original signals to generate a standardized electromagnetic data set;

[0011] A hybrid positioning module is used to perform a joint calculation of the time difference and the angle of arrival on the standardized electromagnetic data set based on a TDOA and AOA fusion positioning algorithm, and output the three-dimensional spatial coordinates of the abnormal radiation source;

[0012] A three-dimensional modeling module is used to associate and map the three-dimensional spatial coordinates with the electromagnetic field intensity data, generate a three-dimensional electromagnetic field intensity distribution map using a spatial interpolation algorithm, and mark the spatial position of the abnormal radiation source;

[0013] The multimodal fusion module is used to synchronously acquire visible light image data, establish the spatiotemporal correlation matrix of geographic coordinates and electromagnetic parameters, extract multidimensional fusion features through tensor decomposition algorithm, and generate an overlay analysis model.

[0014] Preferably, the generation of the dynamic spacing adjustment instruction includes:

[0015] Real-time monitoring of the rotation speed and current parameters of the UAV rotor, and calculation of the electromagnetic interference intensity distribution generated by the rotor;

[0016] Determine the minimum safe distance between electromagnetic sensors based on a preset interference intensity threshold;

[0017] The carbon fiber bracket is driven to extend and retract in sections by a stepper motor, and the sensor is dynamically adjusted to above the minimum safety distance.

[0018] Preferably, the anti-interference filtering process includes:

[0019] Identify the periodic rotor interference waveform in the original signal and construct an adaptive notch filter;

[0020] Wavelet transform algorithm is used to separate high-frequency noise and low-frequency electromagnetic signals;

[0021] The filtered signal is amplitude normalized to generate standardized electromagnetic data with RMS calibration.

[0022] Preferably, the TDOA and AOA fusion positioning algorithm includes:

[0023] Extract the arrival time difference of electromagnetic signals and the phase difference data of multi-antenna arrays to construct an overdetermined set of equations;

[0024] Iteratively solving the overdetermined equations using a weighted least squares method to obtain initial coordinates of the radiation source;

[0025] The initial coordinates are smoothed in time and space by the Kalman filter algorithm to optimize the positioning accuracy.

[0026] Preferably, the spatial interpolation algorithm includes:

[0027] Mapping electromagnetic intensity data of discrete acquisition points to three-dimensional grid nodes;

[0028] The radial basis function interpolation method is used to calculate the field strength value of the unsampled area;

[0029] The interpolation results are corrected by combining the terrain elevation data to generate a continuous three-dimensional field intensity distribution surface.

[0030] Preferably, the construction of the spatiotemporal correlation matrix includes:

[0031] Perform georeferencing on visible light images and extract the conversion relationship between pixel coordinates and longitude and latitude;

[0032] Align the electromagnetic field intensity data with the corresponding geographic coordinates by time stamp;

[0033] Construct a three-dimensional tensor matrix containing spatial position, field strength value and image characteristics.

[0034] Preferably, the extraction of the multi-dimensional fusion features includes:

[0035] Performing high-order singular value decomposition on the spatiotemporal correlation matrix to extract principal component eigenvectors;

[0036] Encode terrain features in visible light images using convolutional neural networks;

[0037] The electromagnetic principal component features and terrain coding features are weightedly fused to generate a joint feature vector.

[0038] Preferably, the system further comprises:

[0039] Construct a radiation source risk assessment model and calculate the regional safety level based on the electromagnetic field intensity and radiation source coordinates;

[0040] Generate a visual map with security level markings and transmit it to the ground control terminal in real time through the wireless communication module.

[0041] Preferably, the system further comprises:

[0042] An electromagnetic field gradient tracking algorithm is embedded in the UAV flight path planning to dynamically adjust the trajectory to focus on high field strength areas.

[0043] Preferably, the present invention further includes an application method based on the above-mentioned multimodal electromagnetic field distribution mapping UAV system, the method comprising:

[0044] Step S1: According to the preset rotor electromagnetic interference suppression rules, the distance between the electromagnetic sensor and the UAV rotor is dynamically adjusted through the retractable carbon fiber bracket, and a dynamic distance adjustment instruction is generated and executed;

[0045] Step S2: Controlling the high-precision three-axis electromagnetic field sensor to collect electromagnetic field signals within a preset frequency band, performing anti-interference filtering and frequency domain conversion processing on the original signal to generate a standardized electromagnetic data set;

[0046] Step S3: Based on the TDOA and AOA fusion positioning algorithm, the time difference measurement and arrival angle joint solution are performed on the standardized electromagnetic data set, and the three-dimensional spatial coordinates of the abnormal radiation source are output after iterative optimization;

[0047] Step S4: performing spatial correlation mapping between the three-dimensional spatial coordinates and the corresponding electromagnetic field intensity data, reconstructing the three-dimensional electromagnetic field intensity distribution surface using the Kriging interpolation algorithm, and marking the spatial position information of the abnormal radiation source;

[0048] Step S5: Synchronously acquire visible light image data, construct a spatiotemporal correlation matrix of geographic coordinates, electromagnetic parameters, and timestamps, extract multidimensional fusion features through high-order singular value tensor decomposition, and generate an electromagnetic-optical superposition analysis model.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The system demonstrates superior performance in suppressing rotor electromagnetic interference (EMI), thanks to its dynamic adjustment module. This module monitors the UAV's rotor speed and current parameters in real time, accurately calculating the distribution of EMI intensity generated by the rotor. A preset interference intensity threshold is used to determine the minimum safe distance between the electromagnetic sensors. A stepper motor then drives the carbon fiber bracket to extend and retract in sections, dynamically adjusting the sensors to a distance above the minimum safe distance. This approach effectively mitigates the impact of rotor EMI on sensor measurement accuracy, ensuring the accuracy and reliability of collected EMI data. Compared to traditional methods, this approach goes beyond simply attempting to mitigate interference, but instead addresses the impact of interference on measurements at the source. This significantly improves measurement data accuracy and provides a solid foundation for subsequent analysis and processing. For example, when conducting electromagnetic field mapping near power facilities, traditional methods can lead to significant measurement deviations due to rotor EMI, making it difficult to accurately determine electromagnetic leakage within the facility. This new approach effectively addresses this issue, accurately measuring EMI intensity and promptly identifying potential safety hazards.

[0051] Another significant advantage of this system is the improved accuracy of locating radiation sources. The hybrid positioning module utilizes a TDOA and AOA fusion positioning algorithm. This algorithm extracts the time difference of arrival of electromagnetic signals and the phase difference data of a multi-antenna array to construct an overdetermined system of equations. The weighted least squares method is then used to iteratively solve the initial coordinates of the radiation source. The Kalman filter algorithm then smooths these initial coordinates in time and space to further optimize positioning accuracy. This fusion positioning algorithm fully combines the advantages of both TDOA and AOA algorithms, effectively reducing environmental interference with positioning. In practical applications, the three-dimensional spatial coordinates of anomalous radiation sources can be accurately determined, whether in complex urban environments or in mountainous areas with complex terrain. This is crucial for the timely detection and resolution of electromagnetic radiation anomalies. For example, in the detection of excessive radiation from communication base stations, the location of the base station can be quickly located, enabling relevant departments to take timely measures for adjustment or repair.

[0052] Multimodal data fusion analysis is a major highlight of this invention. The multimodal fusion module simultaneously acquires visible light image data, constructs a spatiotemporal correlation matrix between geographic coordinates and electromagnetic parameters, and uses a tensor decomposition algorithm to extract multidimensional fusion features, generating an overlay analysis model. When constructing the spatiotemporal correlation matrix, the visible light image is georeferenced, the conversion relationship between pixel coordinates and longitude and latitude is extracted, and the electromagnetic field intensity data is timestamped with the corresponding geographic coordinates to construct a three-dimensional tensor matrix containing spatial position, field intensity values, and image features. Subsequently, the principal component eigenvectors are extracted through high-order singular value decomposition. A convolutional neural network is used to encode the terrain features in the visible light image. The electromagnetic principal component features are weightedly fused with the terrain encoding features to generate a joint eigenvector. This multimodal data fusion approach enables the system to analyze the electromagnetic environment from multiple dimensions. In urban electromagnetic environment assessments, electromagnetic data can be combined with geographic information, topography, and other information to fully understand the distribution patterns of electromagnetic radiation, providing a scientific basis for urban planning, rationally demarcating electromagnetically sensitive areas, and protecting residents from excessive electromagnetic radiation.

[0053] In addition, the system also has a series of practical expansion functions. It builds a radiation source risk assessment model, calculates the regional safety level based on the electromagnetic field intensity and the coordinates of the radiation source, and generates a visual map with safety level annotations. This map is transmitted to the ground control terminal in real time via the wireless communication module, allowing operators to promptly understand the safety status of the on-site electromagnetic environment and make corresponding decisions. The electromagnetic field gradient tracking algorithm is embedded in the UAV flight path planning, and the track is dynamically adjusted to focus on high-field intensity areas, improving surveying and mapping efficiency and ensuring comprehensive and detailed monitoring of key areas. These functions work together to make the entire system more practical and valuable in the field of electromagnetic field mapping, and can meet the diverse needs of electromagnetic field mapping in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a working principle diagram of the multi-modal electromagnetic field distribution mapping UAV system of the present invention;

[0055] Figure 2 This is the workflow diagram for anti-interference filtering processing;

[0056] Figure 3 This is the workflow diagram of the spatial interpolation algorithm;

[0057] Figure 4 A diagram showing the working principle of flight path planning for drones. DETAILED DESCRIPTION

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0059] See also Figures 1-4 The present invention provides a technical solution: a multi-modal electromagnetic field distribution mapping UAV system, which includes: a dynamic adjustment module, a multi-band acquisition module, a hybrid positioning module, a three-dimensional modeling module and a multi-modal fusion module.

[0060] During system operation, the dynamic adjustment module uses a retractable carbon fiber bracket to adjust the distance between the electromagnetic sensor and the drone's rotor according to preset rotor electromagnetic interference suppression rules. This effectively reduces the impact of rotor electromagnetic interference on sensor data collection, and then generates dynamic distance adjustment instructions, laying the foundation for subsequent accurate data collection.

[0061] The multi-band acquisition module controls the high-precision three-axis electromagnetic field sensor, enabling it to collect electromagnetic field signals within a preset frequency band. After acquiring the raw signal, the module applies anti-interference filtering to it, generating a standardized electromagnetic data set that provides high-quality data support for subsequent analysis and processing.

[0062] The hybrid positioning module processes the standardized electromagnetic data set generated by the multi-band acquisition module based on a TDOA and AOA fusion positioning algorithm. By jointly calculating the time difference and arrival angle of the electromagnetic signal, it ultimately outputs the three-dimensional spatial coordinates of the abnormal radiation source, thereby achieving precise positioning of the abnormal radiation source.

[0063] The 3D modeling module correlates and maps the 3D spatial coordinates output by the hybrid positioning module with the electromagnetic field intensity data. Using a spatial interpolation algorithm, it generates a 3D electromagnetic field intensity distribution map and marks the spatial locations of abnormal radiation sources on the map, visually presenting the distribution of the electromagnetic field.

[0064] The multimodal fusion module synchronously acquires visible light image data, establishes a spatiotemporal correlation matrix between geographic coordinates and electromagnetic parameters, uses a tensor decomposition algorithm to extract multidimensional fusion features, and then generates a superposition analysis model to achieve deep fusion and analysis of multimodal data.

[0065] The present invention will be further described below in conjunction with Examples 1 to 5:

[0066] Example 1:

[0067] This embodiment focuses on the generation process of dynamic spacing adjustment instructions in the dynamic adjustment module. While the drone is operating, the system monitors the rotor speed and current parameters in real time. Assuming the rotor speed is n (unit: revolutions per minute) and the current parameter is I (unit: amperes), a specific algorithm is used to calculate the electromagnetic interference intensity distribution generated by the rotor. Generally speaking, the electromagnetic interference intensity E has a certain functional relationship with the rotor speed n and the current I, which can be expressed as E = k1n + k2I, where k1 and k2 are coefficients determined based on the characteristics of the drone's rotor.

[0068] According to the preset interference intensity threshold E th , determine the minimum safety distance d of the electromagnetic sensor min When the calculated electromagnetic interference intensity E is greater than the interference intensity threshold E th , it indicates that the current sensor is subject to large interference and the distance needs to be adjusted.

[0069] The system uses a stepper motor to drive the carbon fiber bracket to expand and contract in sections. The expansion and contraction amount of each step of the stepper motor is Δd. Assuming that after m steps of adjustment, the distance between the sensor and the rotor reaches d, then d=d0+mΔd, where d0 is the initial distance. During the adjustment process, ensure that d is greater than the minimum safe distance d min , thereby dynamically adjusting the sensor to the appropriate position, reducing electromagnetic interference, and ensuring the accuracy of collected data.

[0070] In practical applications, such as electromagnetic field mapping in urban environments, the drone's surroundings are complex and subject to numerous sources of electromagnetic interference. In these situations, the dynamic adjustment module monitors the rotor's speed and current parameters in real time. When the drone encounters airflow changes that cause the rotor's speed to accelerate, the calculated electromagnetic interference intensity increases. Based on this calculation and adjustment process, the system uses a stepper motor to extend the carbon fiber bracket, increasing the distance between the electromagnetic sensor and the rotor. This successfully avoids areas of strong electromagnetic interference generated by the rotor, making the collected electromagnetic field signals more accurate and reliable, and providing a high-quality data foundation for subsequent analysis.

[0071] Example 2:

[0072] See attached Figure 2 This embodiment details the anti-interference filtering process in the multi-band acquisition module. The anti-interference filtering process includes:

[0073] First, the periodic rotor interference waveform in the original signal is identified, and an adaptive notch filter is constructed based on it.

[0074] The wavelet transform algorithm is used to separate high-frequency noise from low-frequency electromagnetic signals.

[0075] Perform amplitude normalization on the filtered signal.

[0076] Generate standardized electromagnetic data calibrated by the root mean square value, and then check whether the standardized electromagnetic data meets the requirements. If it meets the requirements, the process ends. If it does not meet the requirements, return to the amplitude normalization step of the filtered signal and re-operate.

[0077] After a high-precision three-axis electromagnetic field sensor acquires a raw signal within a preset frequency band, it first identifies the periodic rotor interference waveform within the raw signal. By analyzing the signal characteristics, an adaptive notch filter is constructed. This filter specifically removes the periodic rotor interference signal. Let's assume the raw signal is x(t) and the signal processed by the adaptive notch filter is y1(t). The filter operates based on an adaptive filtering algorithm, which continuously adjusts the filter parameters to minimize the rotor interference component in the output signal y1(t).

[0078] A wavelet transform algorithm is used to separate high-frequency noise from low-frequency electromagnetic signals. Wavelet transform is a time-frequency analysis method that decomposes signals at different frequency scales. Let the high-frequency component after wavelet transform be H(f) and the low-frequency component be L(f). By properly selecting the wavelet basis function and the number of decomposition levels, the high-frequency noise is primarily concentrated in the high-frequency component H(f), while the low-frequency electromagnetic signal is primarily contained in the low-frequency component L(f).

[0079] Perform amplitude normalization on the filtered signal. Assume that the signal after the above two steps is y2(t), and its amplitude is A. In order to facilitate subsequent analysis and comparison, its amplitude needs to be normalized to a certain range. By calculating the root mean square value RMS of the signal and performing calibration, the root mean square value calibrated standardized electromagnetic data is generated. The calculation formula of the root mean square value RMS is , where T is the signal sampling time. After amplitude normalization, the generated standardized electromagnetic data can more intuitively reflect the true characteristics of the electromagnetic field, providing reliable data support for subsequent positioning and analysis.

[0080] For example, when mapping electromagnetic fields near power facilities, the raw signal contains not only the electromagnetic field signals generated by the power facilities, but also interference from drone rotors and high-frequency noise from the surrounding environment. The multi-band acquisition module's anti-interference filtering successfully removes this interference and noise, generating accurate, standardized electromagnetic data. This data plays a key role in subsequently determining the range and intensity distribution of electromagnetic radiation from power facilities, providing a robust basis for assessing their impact on the surrounding environment.

[0081] Example 3:

[0082] This example focuses on the TDOA and AOA fusion positioning algorithm in the hybrid positioning module. First, the arrival time difference of electromagnetic signals and the phase difference data of the multi-antenna array are extracted from the standardized electromagnetic data set. Assuming the arrival times of the electromagnetic signals received by different antennas are t1 and t2, respectively, the arrival time difference Δt = t2 - t1. The phase difference data of the multi-antenna array is Δφ. Based on this data, an overdetermined system of equations is constructed.

[0083] The weighted least squares method is used to iteratively solve the overdetermined equations to obtain the initial coordinates of the radiation source. The goal of the weighted least squares method is to minimize the weighted sum of squares of the errors. Let the error vector be , the weight matrix is W, then the objective function is By continuously iteratively adjusting the coordinate values, the objective function J is minimized, thereby obtaining the initial coordinates (x0, y0, z0) of the radiation source.

[0084] Since the initial coordinates may have certain errors, the Kalman filter algorithm is used to perform spatiotemporal smoothing on the initial coordinates to optimize positioning accuracy. The Kalman filter algorithm is an optimal estimation method based on the state space model. It continuously integrates new measurement data and previous estimation results through two steps: prediction and update, thereby improving positioning accuracy. Let the state vector be X, the measurement vector be Z, and the prediction equation be , the update equation is , where F k is the state transfer matrix, Q k is the process noise covariance matrix, K k is the Kalman gain, H k is the observation matrix. After processing by the Kalman filter algorithm, more accurate three-dimensional spatial coordinates of the abnormal radiation source can be obtained, thereby improving the positioning accuracy of the abnormal radiation source.

[0085] For example, when locating the radiation source of a communication base station, the hybrid positioning module uses collected electromagnetic signal data and a TDOA and AOA fusion positioning algorithm to accurately calculate the three-dimensional spatial coordinates of the communication base station. In cities, where communication base stations are densely distributed, traditional positioning methods are susceptible to interference. However, the fusion positioning algorithm of this invention, by considering multiple factors such as arrival time difference and phase difference, and combining it with a Kalman filter algorithm for optimization, successfully overcomes this interference problem and achieves high-precision positioning of communication base station radiation sources, providing accurate data support for communication network optimization and electromagnetic environment assessment.

[0086] Example 4:

[0087] See attached Figure 3 This embodiment details the spatial interpolation algorithm in the 3D modeling module. The process of the spatial interpolation algorithm includes:

[0088] First, obtain the electromagnetic intensity data of discrete collection points.

[0089] Map the electromagnetic intensity data of discrete acquisition points to three-dimensional grid nodes.

[0090] The radial basis function interpolation method is used to calculate the field strength value of the unsampled area.

[0091] Obtain terrain elevation data and perform boundary correction on the interpolation results based on the terrain elevation data.

[0092] Determine whether the boundary correction meets the requirements. If so, generate a continuous three-dimensional field intensity distribution surface. If not, return to the boundary correction step and reprocess the interpolation result in combination with the terrain elevation data.

[0093] When performing 3D modeling, the electromagnetic intensity data of discrete acquisition points are first mapped to 3D grid nodes. Let the electromagnetic intensity data of the acquisition point be E i (i=1,2,…,n), the coordinates of the three-dimensional grid nodes are (x j ,y j ,z j )(j=1,2,…,m), the collected point data are assigned to the corresponding grid nodes through specific mapping rules.

[0094] The radial basis function interpolation method is used to calculate the field strength value of the unsampled area. The radial basis function interpolation method is a commonly used interpolation method. Let the radial basis function be φ(r), where r is the distance. Generally, Gaussian function can be selected as the radial basis function. For the unsampled point (x, y, z), the calculation formula of its field strength value E(x, y, z) is , where λ i is the weight coefficient, which is determined by solving the linear equations.

[0095] The interpolated results are then bounded and corrected using terrain elevation data to generate a continuous three-dimensional field intensity distribution surface. Assuming the terrain elevation data is h(x,y), the interpolated field intensity values are corrected to account for the impact of terrain on electromagnetic field propagation. For example, in areas with complex terrain, such as mountainous areas, electromagnetic field propagation is affected by the undulating terrain. By incorporating terrain elevation data, the field intensity values in areas near mountains are adjusted, making the generated three-dimensional field intensity distribution surface more realistic and accurately reflecting the distribution characteristics of the electromagnetic field under different terrain conditions.

[0096] For example, when mapping the electromagnetic field distribution of power transmission lines in mountainous areas, the 3D modeling module uses a spatial interpolation algorithm to convert discrete data points into a continuous 3D field intensity distribution surface. Taking into account the complex terrain of mountainous areas, boundary corrections are performed by incorporating terrain elevation data, successfully demonstrating how the electromagnetic field varies across different terrain areas, such as valleys and hillsides. This provides an intuitive and accurate basis for assessing the impact of power transmission lines on the ecological environment and residents' lives in mountainous areas, and facilitates the development of appropriate protective measures and planning schemes.

[0097] Example 5:

[0098] This embodiment mainly introduces the construction of the spatiotemporal correlation matrix in the multimodal fusion module, the extraction of multidimensional fusion features, and the implementation of other functions of the system.

[0099] When constructing the spatiotemporal correlation matrix, the visible light imagery is first georeferenced to extract the conversion relationship between pixel coordinates and longitude and latitude. Let the pixel coordinates be (u, v) and the longitude and latitude coordinates be (φ, λ). Using the georeference algorithm, we obtain the conversion formula (φ, λ) = f(u, v), where f is the conversion function.

[0100] Align the electromagnetic field intensity data with the corresponding geographic coordinates by timestamp. Let the electromagnetic field intensity data be E(t), the geographic coordinates be (φ(t), λ(t)), and the timestamp be t. Ensure that the electromagnetic field intensity data at each moment is accurately associated with the corresponding geographic coordinates.

[0101] A three-dimensional tensor matrix containing spatial position, field strength, and image features is constructed. This tensor matrix fuses multimodal data together and provides a basis for subsequent analysis.

[0102] When extracting multidimensional fusion features, we perform high-order singular value decomposition (HSVD) on the spatiotemporal correlation matrix to extract the principal component eigenvectors. HSVD decomposes the tensor matrix into the product of multiple low-dimensional matrices, thereby extracting the key feature information. Let T be the spatiotemporal correlation matrix. After HSVD, we obtain the principal component eigenvectors U1, U2, U3, etc.

[0103] The terrain features in visible light imagery are encoded using a convolutional neural network. Convolutional neural networks have powerful feature extraction capabilities and can automatically learn the terrain features in images. Let the output of the convolutional neural network be F, which contains rich terrain feature information.

[0104] The electromagnetic principal component features and terrain coding features are weighted and fused to generate a joint feature vector. Let the electromagnetic principal component feature vector be E, the terrain coding feature vector be T, and the weighting coefficients be ω1 and ω2 respectively, then the joint feature vector J = ω1E + ω2T.

[0105] The system also constructs a radiation source risk assessment model to calculate the regional safety level based on the electromagnetic field intensity and the radiation source coordinates. Assuming the electromagnetic field intensity E and the radiation source coordinates (x, y, z), the regional safety level R is calculated using the risk assessment model R = g(E, x, y, z), where g is the risk assessment function. A visual map with safety level annotations is generated and transmitted in real time to the ground control terminal via a wireless communication module, allowing operators to promptly understand the electromagnetic field distribution and safety status of the surveyed area.

[0106] See attached Figure 4 , is the flow chart of UAV flight path planning. System initialization, planning the initial flight path of the UAV, embedding the electromagnetic field gradient tracking algorithm in the UAV flight path planning. During the flight, the electromagnetic field intensity data is monitored, and the electromagnetic field gradient tracking algorithm is used to determine whether to adjust the trajectory: if so, the trajectory is dynamically adjusted to focus on the high field intensity area; otherwise, the original trajectory is flown. Let the electromagnetic field gradient be ,According to the direction and size of the electromagnetic field gradient, the flight direction and speed of the UAV are adjusted, so that the UAV can fly towards the high field strength area, improving the surveying and mapping efficiency and accuracy.

[0107] For example, when conducting multimodal electromagnetic field distribution mapping in a chemical park, the multimodal fusion module constructs a spatiotemporal correlation matrix and extracts multidimensional fusion features, enabling a comprehensive analysis of the park's electromagnetic environment and topography. Simultaneously, it utilizes radiation source risk assessment models and visualization maps to promptly identify areas of potential electromagnetic safety hazards. Furthermore, the electromagnetic field gradient tracking algorithm incorporated into drone flight path planning guides drones to conduct detailed mapping of key areas, providing comprehensive and accurate data support for chemical park safety management and environmental assessments.

[0108] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0109] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multi-modal electromagnetic field distribution mapping UAV system, characterized in that: The system comprises: The dynamic adjustment module is used to adjust the distance between the electromagnetic sensor and the UAV rotor through the retractable carbon fiber bracket according to the preset rotor electromagnetic interference suppression rules and generate dynamic distance adjustment instructions; The multi-band acquisition module is used to control the high-precision three-axis electromagnetic field sensor to collect electromagnetic field signals within a preset frequency band, and perform anti-interference filtering on the collected original signals to generate a standardized electromagnetic data set; A hybrid positioning module is used to perform a joint calculation of the time difference and the angle of arrival on the standardized electromagnetic data set based on a TDOA and AOA fusion positioning algorithm, and output the three-dimensional spatial coordinates of the abnormal radiation source; A three-dimensional modeling module is used to associate and map the three-dimensional spatial coordinates with the electromagnetic field intensity data, generate a three-dimensional electromagnetic field intensity distribution map using a spatial interpolation algorithm, and mark the spatial position of the abnormal radiation source; The multimodal fusion module is used to synchronously acquire visible light image data, establish the spatiotemporal correlation matrix of geographic coordinates and electromagnetic parameters, extract multidimensional fusion features through tensor decomposition algorithm, and generate an overlay analysis model.

2. The multimodal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The generation of the dynamic spacing adjustment instruction includes: Real-time monitoring of the rotation speed and current parameters of the UAV rotor, and calculation of the electromagnetic interference intensity distribution generated by the rotor; Determine the minimum safe distance between electromagnetic sensors based on a preset interference intensity threshold; The carbon fiber bracket is driven to extend and retract in sections by a stepper motor, and the sensor is dynamically adjusted to above the minimum safety distance.

3. The multimodal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The anti-interference filtering process includes: Identify the periodic rotor interference waveform in the original signal and construct an adaptive notch filter; Wavelet transform algorithm is used to separate high-frequency noise and low-frequency electromagnetic signals; The filtered signal is amplitude normalized to generate standardized electromagnetic data with RMS calibration.

4. The multi-modal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The TDOA and AOA fusion positioning algorithm includes: Extract the arrival time difference of electromagnetic signals and the phase difference data of multi-antenna arrays to construct an overdetermined set of equations; Iteratively solving the overdetermined equations using a weighted least squares method to obtain initial coordinates of the radiation source; The initial coordinates are smoothed in time and space by the Kalman filter algorithm to optimize the positioning accuracy.

5. The multi-modal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The spatial interpolation algorithm includes: Mapping electromagnetic intensity data of discrete acquisition points to three-dimensional grid nodes; The radial basis function interpolation method is used to calculate the field strength value of the unsampled area; The interpolation results are corrected by combining the terrain elevation data to generate a continuous three-dimensional field intensity distribution surface.

6. The multi-modal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The construction of the spatiotemporal correlation matrix includes: Perform georeferencing on visible light images and extract the conversion relationship between pixel coordinates and longitude and latitude; Align the electromagnetic field intensity data with the corresponding geographic coordinates by time stamp; Construct a three-dimensional tensor matrix containing spatial position, field strength value and image characteristics.

7. The multi-modal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The extraction of the multi-dimensional fusion features includes: Performing high-order singular value decomposition on the spatiotemporal correlation matrix to extract principal component eigenvectors; Encode terrain features in visible light images using convolutional neural networks; The electromagnetic principal component features and terrain coding features are weightedly fused to generate a joint feature vector.

8. The multi-modal electromagnetic field distribution mapping UAV system according to claim 1, characterized in that: The system further comprises: Construct a radiation source risk assessment model and calculate the regional safety level based on the electromagnetic field intensity and radiation source coordinates; Generate a visual map with security level markings and transmit it to the ground control terminal in real time through the wireless communication module.

9. The multi-modal electromagnetic field distribution mapping UAV system according to any one of claims 1 to 8, characterized in that: The system further comprises: An electromagnetic field gradient tracking algorithm is embedded in the UAV flight path planning to dynamically adjust the trajectory to focus on high field strength areas.

10. An application method of the multi-modal electromagnetic field distribution mapping UAV system according to any one of claims 1 to 9, characterized in that: include: Step S1: According to the preset rotor electromagnetic interference suppression rules, the distance between the electromagnetic sensor and the UAV rotor is dynamically adjusted through the retractable carbon fiber bracket, and a dynamic distance adjustment instruction is generated and executed; Step S2: Controlling the high-precision three-axis electromagnetic field sensor to collect electromagnetic field signals within a preset frequency band, performing anti-interference filtering and frequency domain conversion processing on the original signal to generate a standardized electromagnetic data set; Step S3: Based on the TDOA and AOA fusion positioning algorithm, the time difference measurement and arrival angle joint solution are performed on the standardized electromagnetic data set, and the three-dimensional spatial coordinates of the abnormal radiation source are output after iterative optimization; Step S4: performing spatial correlation mapping between the three-dimensional spatial coordinates and the corresponding electromagnetic field intensity data, reconstructing the three-dimensional electromagnetic field intensity distribution surface using the Kriging interpolation algorithm, and marking the spatial position information of the abnormal radiation source; Step S5: Synchronously acquire visible light image data, construct a spatiotemporal correlation matrix of geographic coordinates, electromagnetic parameters, and timestamps, extract multidimensional fusion features through high-order singular value tensor decomposition, and generate an electromagnetic-optical superposition analysis model.

Images