Dam termite cave positioning method based on unmanned aerial vehicle carrying transient electromagnetism

Abstract

The invention discloses a dam termite cave positioning method based on transient electromagnetism carried by an unmanned aerial vehicle. The method comprises the following steps: constructing an unmanned aerial vehicle transient electromagnetic detection system; preprocessing the received secondary induction electromagnetic field signal, and obtaining a comprehensive electromagnetic signal according to the response difference of the multi-frequency signal; inputting the integrated electromagnetic signals into a pre-trained deep learning neural network model, identifying characteristic signals related to termite acupoints in the integrated electromagnetic signals, and outputting position coordinates of suspected termite acupoints; a dam surface thermal imaging image obtained by a thermal imaging module is combined for secondary verification, and termite cave position information is transmitted to a ground control terminal in real time through a communication module carried by the unmanned aerial vehicle. According to the invention, efficient, accurate and lossless positioning of the termite cave of the dam is realized, underground electromagnetic signals and surface heat distribution information of the dam are rapidly obtained, and accurate termite cave position coordinates are output.

Description

Technical Field

[0001] The present invention relates to the technical field of safety monitoring of water conservancy projects, and particularly to a method for locating termite nests in dams based on unmanned aerial vehicle (UAV)-borne transient electromagnetic method. Background Technique

[0002] In the field of water conservancy projects, dams, as key facilities for flood control and water storage, their safety is directly related to the safety of people's lives and property and the ecological environment. Termites pose a great threat to dams. Termites build nests and construct ant galleries in dams, which will form a hollow structure, reduce the strength of the dams, and even cause serious accidents such as piping and dam break. According to statistics, the proportion of dam safety hazards caused by termite damage has been increasing year by year. Therefore, developing efficient and accurate termite nest location technology in dams is of great significance for ensuring the safe operation of dams.

[0003] Traditional termite nest location technology in dams mainly relies on manual inspection. Technical personnel judge the location of termite nests by observing obvious signs such as ant roads and swarming holes on the dam surface and combining the method of listening to the sound difference by knocking on the dam surface. This method depends on the experience of inspectors, has strong subjectivity, and there are large differences in judgment criteria and capabilities among different personnel, which is prone to misjudgment or missed judgment. At the same time, the efficiency of manual inspection is extremely low. Facing long-distance and large-area dams, a large amount of manpower and time are required. Moreover, in areas with complex terrain and difficult access, such as steep slopes and underwater areas, it is difficult to effectively carry out manual inspection. In addition, manual inspection can only detect surface or shallow termite activity signs and cannot accurately detect deep main nests and complex ant gallery systems. Although the cost of manual inspection is relatively low and the operation is simple, due to its disadvantages such as poor accuracy, low efficiency, and limited detection range, it is difficult to meet the requirements of modern dam safety monitoring.

[0004] With the development of technology, some existing technologies have begun to adopt physical detection means, such as ground penetrating radar and resistivity method. Ground penetrating radar emits high-frequency electromagnetic waves, receives the reflected waves from underground media, and judges the position of the target object according to the echo time and waveform characteristics. This method has high resolution, can quickly obtain underground structure information, and has no damage to the dam. However, ground penetrating radar is greatly affected by factors such as soil moisture content and dielectric constant of the medium. In dam areas with high water content, the electromagnetic wave attenuation is serious, and the detection depth and accuracy are significantly reduced. At the same time, ground penetrating radar has poor recognition effect on termite nests with irregular shapes and small sizes, and is prone to misjudging other geological anomalies as termite nests.

[0005] The resistivity method utilizes the resistivity differences of different underground media to detect the target object. By arranging electrodes on the ground, measuring the underground current and potential distributions, and inversely calculating the underground resistivity distribution image. This method is suitable for large-area detection and has a certain detection ability for deep target objects. However, the resistivity method is susceptible to electromagnetic interference from the surrounding environment, with complex data processing and strong multi-solution characteristics in the inversion results. It is difficult to accurately distinguish the resistivity differences between termite nests and other geological structures, resulting in relatively low positioning accuracy. Moreover, this method requires a large number of electrodes to be arranged on the ground, with cumbersome operations and poor adaptability to complex terrains, unable to meet the requirements of rapid and accurate detection of dams. Summary of the Invention

[0006] Based on the above technical problems, the present application discloses a method for locating termite nests in dams based on unmanned aerial vehicle (UAV)-borne transient electromagnetic method, including:

[0007] Construct a UAV transient electromagnetic detection system;

[0008] Preprocess the received secondary induced electromagnetic field signals, and calculate the weights of signals with different frequencies according to the response differences of multi-frequency signals, and fuse them to obtain a comprehensive electromagnetic signal;

[0009] Input the comprehensive electromagnetic signal into a pre-trained deep learning neural network model, identify the characteristic signals related to termite nests in the comprehensive electromagnetic signal, and output the position coordinates of suspected termite nests;

[0010] Combined with the thermal imaging image of the dam surface obtained by the thermal imaging module, conduct a secondary verification on the position coordinates of suspected termite nests output by the deep learning neural network model. If there is an abnormal temperature distribution area at the corresponding position in the thermal imaging image, determine that position as the termite nest position, and transmit the termite nest position information to the ground control terminal in real time through the communication module carried by the UAV.

[0011] Preferably, the UAV transient electromagnetic detection system adopts a multi-frequency time-division emission technology, emitting at least three different frequencies of pulsed current signals; multiple independent emission waveform generation modules are arranged in the transient electromagnetic detection system, and each module corresponds to a pulsed current signal with a different frequency. The programmable logic controller is used to control the timing switching circuit to sequentially activate each emission waveform generation module at a preset time interval; within each emission cycle, first emit a low-frequency pulsed current signal for detecting the deep area of the dam, emit a medium-frequency pulsed current signal after an interval of 5 - 10 milliseconds to detect the middle area, and then emit a high-frequency pulsed current signal after an interval of 3 - 5 milliseconds to detect the shallow area. The emission duration and current intensity of each frequency signal are adaptively adjusted according to the geological conditions of the dam and historical detection data to obtain effective excitation and detection in different depth areas.

[0012] Preferably, the UAV transient electromagnetic detection system further includes an inertial navigation module, a lidar altimetry module, and a thermal imaging module; the inertial navigation module is used to obtain the attitude and position information of the UAV in real time; the lidar altimetry module is used to monitor the distance between the UAV and the dam surface in real time; the thermal imaging module is used to assist in identifying abnormal heat source areas on the dam surface.

[0013] Preferably, the preprocessing of the secondary induced electromagnetic field signal is specifically as follows: the original signal collected is processed by a wavelet transform denoising algorithm and decomposed into high-frequency components and low-frequency components at different scales The formula is: where is the number of wavelet decomposition layers. An adaptive threshold is set for the high-frequency components , and an improved soft threshold function is used to denoise each layer of high-frequency components to generate denoised high-frequency components . The formula is:

[0014]

[0015] where is the sign function. The denoised high-frequency components and the low-frequency components are reconstructed through wavelet inverse transform to obtain the denoised signal .

[0016] Preferably, based on the denoised signal, according to the response differences of multi-frequency signals, the weights of different frequency signals are calculated . The formula is: where is the standard deviation of the th frequency signal, is the number of frequency signals; through weighted summation of each frequency signal is performed to obtain a comprehensive electromagnetic signal . The formula is: where is the denoised signal of the th frequency.

[0017] Preferably, the pre-trained deep learning neural network model adopts a dual-channel attention convolutional neural network structure, including parallel frequency feature channels and time series channels; the frequency feature channels adopt a multi-layer grouped convolutional module, and the convolutional kernel size and dilation rate of each group of convolutional layers are adaptively adjusted according to the characteristics of multi-frequency signals to extract the features of electromagnetic signals with different frequencies; the time series channels adopt a structure combining long short-term memory network and self-attention mechanism to capture the sequence features of electromagnetic signals changing over time; a cross-channel attention module is set after the two channels, and by calculating the mutual information between the frequency features and the time series features, a channel-intermediate attention weight matrix is generated , the formula is , where is the frequency feature vector, is the time series feature vector, and are linear transformation functions respectively; the weighted features of the two channels are cascaded and then input into the global average pooling layer and the fully connected layer, and the focal loss function is used for training, where and are hyperparameters, is the probability of termite nest predicted by the model; the training data includes the electromagnetic signal data of the dam under different geological conditions and different termite damage degrees and the corresponding annotation information of the actual location of the termite nest, enhancing the recognition ability of the characteristic signals of the termite nest in the dam

[0018] Preferably, the process of identifying the characteristic signals related to the termite nest in the comprehensive electromagnetic signal by the deep learning neural network model and outputting the suspected termite nest position coordinates is as follows: the comprehensive electromagnetic signal is vectorized according to the time series and frequency dimensions and converted into a three-dimensional data tensor , where is the number of time sampling points, is the number of frequency channels, is the number of data channels; the data tensor is input into the deep learning neural network model, the grouped convolutional layers of the frequency feature channels extract the local electromagnetic features at different frequencies, the LSTM-self-attention module of the time series channels captures the changing law of the electromagnetic signals over time, and then the cross-channel attention module performs weight allocation and fusion on the features of the two channels; at the model output layer, through the position decoding mechanism of the conditional random field, a joint probability model of the characteristic signals and the spatial positions is constructed, and the formula is: , where is the position state sequence, is the normalization factor, is the transfer potential energy function of adjacent positions, is the characteristic potential energy function of the current position is the position state variable corresponding to the moment in the time series. By performing an optimal path search on the joint probability model, the position states with probability values exceeding the preset threshold are mapped to coordinates in the geographic coordinate system, and the position coordinates of the suspected termite nests are output.

[0019] Preferably, the thermal imaging module obtains the thermal imaging image of the dam surface. Specifically: the thermal imaging module uses a dual-band infrared detector to simultaneously collect the thermal radiation signals in the medium-wave infrared band of 3-5μm and the long-wave infrared band of 8-14μm. During the collection process, a galvanometer scanning mechanism is used to perform point-by-point scanning on the dam surface in a spiral scanning path, and the scanning angle range covers the ±60° field of view directly below the drone; during the scanning process, combined with the distance information obtained in real time by the lidar altimetry module, geometric distortion correction is performed on the thermal imaging image. Through the background suppression algorithm of adaptive Kalman filtering, a background temperature dynamic model is established, and the formula is: , where is the predicted background temperature at the current moment, is the predicted value at the previous moment, is the adaptive Kalman gain, is the currently collected temperature value. To reduce the influence of environmental temperature changes on imaging, the dual-band thermal radiation signals are pixel-level fused, and the formula is: , generating the final thermal imaging image, where and are the medium-wave and long-wave infrared images respectively, , are the corresponding dynamic adjustment coefficients.

[0020] Preferably, the process of secondarily verifying the position coordinates of the suspected termite nests by combining the thermal imaging image of the dam surface obtained by the thermal imaging module is as follows. The position coordinates of the suspected termite nests output by the deep learning neural network model are converted into the pixel coordinate system of the thermal imaging image. The mapping relationship between the geographic coordinate system and the image coordinate system is established through a coordinate transformation matrix. With the converted pixel coordinates as the center, a local area of pixel size is delimited, and the temperature distribution data within the area is extracted. Through the abnormal temperature area recognition algorithm of density peak clustering, the local density and distance of each pixel point within the area are calculated. The pixel points with the product of the local density and distance exceeding the set threshold are clustered into abnormal temperature areas. If there is such an abnormal temperature clustering area within the delimited local area, it is determined that the suspected position is the termite nest position; if not, the suspected position is excluded, and the result is fed back to the ground control terminal.

[0021] ​Preferably, the communication module adopts the ultra-wideband wireless transmission technology in the time-division duplex mode to construct a dam local Internet of Things. The position information data frame is encapsulated by the Beidou short message protocol and directionally transmitted by the MIMO antenna array after channel coding, realizing the real-time transmission and anti-interference transmission of centimeter-level positioning data.

[0022] Compared with the prior art, the technical solution of the present application has the following technical effects:

[0023] The present invention constructs a detection system by carrying a transient electromagnetic device and multiple sensors on an unmanned aerial vehicle (UAV), and automatically flies in combination with a preset route, which can quickly cover a large area of the dam area. Compared with the problem of low efficiency in traditional manual inspection that requires piecemeal inspection, the UAV can complete the detection task of a long-distance dam in a short time, greatly shortening the detection cycle. At the same time, the multi-frequency time-division emission technology cooperates with multi-sensor data acquisition to obtain multi-dimensional detection information at one time, reducing repeated operations and significantly improving the overall working efficiency of termite nest location in the dam, meeting the requirements of modern water conservancy projects for efficient monitoring.

[0024] The present invention utilizes the sensitive characteristics of the transient electromagnetic method to the conductivity difference of underground media, and combines the feature recognition of the comprehensive electromagnetic signal by the deep learning neural network model to accurately capture the electromagnetic response difference between the termite nest and the surrounding soil, effectively avoiding misjudgment caused by interference factors such as geological structure and underground pipelines. In addition, the secondary verification mechanism of the thermal imaging module further confirms the suspected position from the perspective of temperature distribution. Through dual-band imaging, geometric correction and abnormal temperature area recognition algorithms, the reliability of the detection results is enhanced, realizing high-precision positioning of the termite nest location and providing an accurate basis for subsequent prevention and control work.

[0025] The present invention adopts a non-contact detection method, and the UAV does not need to directly contact the dam, avoiding the damage to the dam structure caused by excavation or drilling in the traditional detection method, ensuring the safety and integrity of the dam. At the same time, the flexible mobility of the UAV enables it to adapt to various complex terrain environments. Whether it is a mountainous steep slope, a dam surrounded by water, or an area difficult for humans to reach, the detection work can be carried out smoothly. In addition, the communication module realizes real-time data transmission, facilitating remote monitoring and decision-making, making the technology have a wide range of application scenarios and environmental adaptability.

[0026] The above description is only an overview of the technical solution of the present application. In order to be able to understand the technical means of the present application more clearly, so as to be implemented in accordance with the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following describes in detail with the preferred embodiments of the present application and in conjunction with the accompanying drawings.

[0027] Those skilled in the art will better understand the above and other objects, advantages and features of the present application from the following detailed description of specific embodiments of the present application in conjunction with the accompanying drawings. Description of the Drawings

[0028] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0029] Figure 1 Flowchart of the method for locating termite nests in dams based on transient electromagnetic carried by unmanned aerial vehicles in the present invention;

[0030] Figure 2 Structural diagram of the dual-channel attention convolutional neural network in the present invention;

[0031] Figure 3 Comparison chart of the accuracy rate between the transient electromagnetic in the present invention and the prior art;

[0032] Figure 4 Comparison chart of the misjudgment rate between the transient electromagnetic in the present invention and the prior art;

[0033] Figure 5 Comparison chart of the convergence rate between the deep learning neural network and the traditional convolutional neural network in the present invention. Detailed Embodiments

[0034] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. In the following description, specific details such as specific configurations and components are provided only to assist in a comprehensive understanding of the embodiments of the present application. Therefore, those skilled in the art should clearly understand that various changes and modifications can be made to the embodiments described here without departing from the scope and spirit of the present application. In addition, descriptions of known functions and structures are omitted for clarity and conciseness in the embodiments.

[0035] It should be understood that the "one embodiment" or "this embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, the "one embodiment" or "this embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner.

[0036] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not itself indicate the relationship between the various embodiments and / or arrangements discussed.

[0037] The term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" in this article is a description of another association object relationship, indicating that there can be two relationships. For example, A / and B can represent: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0038] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there can be three relationships. For example, at least one of A and B can represent: A exists alone, A and B exist simultaneously, and B exists alone.

[0039] It should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion.

[0040] Embodiment 1

[0041] This embodiment mainly describes a method for locating termite nests in dikes based on transient electromagnetic detection carried by an unmanned aerial vehicle (UAV), as Figure 1 shown, specifically including:

[0042] Construct a UAV transient electromagnetic detection system;

[0043] Preprocess the received secondary induced electromagnetic field signals, and calculate the weights of signals with different frequencies according to the response differences of multi-frequency signals, and fuse to obtain a comprehensive electromagnetic signal;

[0044] Input the comprehensive electromagnetic signal into a pre-trained deep learning neural network model to identify the feature signals related to termite nests in the comprehensive electromagnetic signal and output the position coordinates of suspected termite nests.

[0045] Combine the thermal imaging image of the dam surface obtained by the thermal imaging module to perform secondary verification on the position coordinates of the suspected termite nests output by the deep learning neural network model. If there is an abnormal temperature distribution area at the corresponding position in the thermal imaging image, determine that position as the termite nest position and transmit the termite nest position information to the ground control terminal in real time through the communication module carried by the drone.

[0046] Furthermore, the drone transient electromagnetic detection system adopts a multi-frequency time-sharing emission technology to emit at least three different frequency pulsed current signals; multiple independent emission waveform generation modules are set in the transient electromagnetic detection system, and each module corresponds to a pulsed current signal of a different frequency. The programmable logic controller is used to control the timing switching circuit to sequentially activate each emission waveform generation module at a preset time interval; within each emission cycle, first emit a low-frequency pulsed current signal to detect the deep area of the dam, emit a medium-frequency pulsed current signal after an interval of 5 - 10 milliseconds to detect the middle area, and then emit a high-frequency pulsed current signal after an interval of 3 - 5 milliseconds to detect the shallow area. The emission duration and current intensity of each frequency signal are adaptively adjusted according to the geological conditions of the dam and historical detection data to obtain effective excitation and detection in different depth areas.

[0047] Furthermore, the drone transient electromagnetic detection system also includes an inertial navigation module, a lidar altimetry module, and a thermal imaging module; the inertial navigation module is used to obtain the attitude and position information of the drone in real time; the lidar altimetry module is used to monitor the distance between the drone and the dam surface in real time; the thermal imaging module is used to assist in identifying abnormal heat source areas on the dam surface.

[0048] Furthermore, the preprocessing of the secondary induced electromagnetic field signal is specifically as follows: process the collected original signal through the wavelet transform denoising algorithm, decompose it into high-frequency components and low-frequency components , and the formula is: , where is the number of wavelet decomposition layers. Set an adaptive threshold for the high-frequency components , and perform denoising processing on each layer of high-frequency components using an improved soft threshold function to generate the denoised high-frequency components , and the formula is:

[0049]

[0050] where is the sign function, and the denoised high-frequency component and the low-frequency component are reconstructed through inverse wavelet transform to obtain the denoised signal .

[0051] Furthermore, based on the denoised signal, according to the response differences of multi-frequency signals, the weights of signals with different frequencies are calculated , and the formula is: , where is the standard deviation of the th frequency signal, is the number of frequency signals; through weighted summation of each frequency signal is performed to obtain the comprehensive electromagnetic signal , and the formula is: , where is the denoised signal of the th frequency.

[0052] Furthermore, as Figure 2 shown, the pre-trained deep learning neural network model adopts a dual-channel attention convolutional neural network structure, including parallel frequency feature channels and time series channels; the frequency feature channels adopt a multi-layer grouped convolutional module, and the convolutional kernel size and dilation rate of each group of convolutional layers are adaptively adjusted according to the characteristics of multi-frequency signals for extracting the features of electromagnetic signals with different frequencies; the time series channels adopt a structure combining long short-term memory network and self-attention mechanism to capture the sequential features of electromagnetic signals changing over time; a cross-channel attention module is set after the two channels, and by calculating the mutual information between the frequency feature and the time series feature, a channel-inter attention weight matrix is generated, and the formula is , where is the frequency feature vector, is the time series feature vector, and are linear transformation functions respectively; the weighted features of the two channels are concatenated and then input into the global average pooling layer and the fully connected layer, and the focal loss function is used for training, where and are hyperparameters, is the probability of termite nests predicted by the model; the training data includes the electromagnetic signal data of the dam under different geological conditions and different termite damage degrees and the corresponding annotation information of the actual positions of termite nests, enhancing the recognition ability of the characteristic signals of termite nests in the dam.

[0053] Furthermore, the process of identifying the characteristic signals related to termite nests in the comprehensive electromagnetic signal through the deep learning neural network model and outputting the suspected termite nest position coordinates is as follows: the comprehensive electromagnetic signal Vectorize according to the time series and frequency dimension and convert it into a three-dimensional data tensor , where is the number of time sampling points, is the number of frequency channels, is the number of data channels; input the data tensor into the deep learning neural network model. Extract the local electromagnetic features at different frequencies through the grouped convolutional layer of the frequency feature channel, capture the changing law of the electromagnetic signal over time by the LSTM-self-attention module of the time series channel, and then perform weight allocation and fusion on the features of the two channels through the cross-channel attention module; at the model output layer, construct a joint probability model of the feature signal and the spatial position through the position decoding mechanism of the conditional random field. The formula is: , where is the position state sequence, is the normalization factor, is the transfer potential energy function of adjacent positions, is the feature potential energy function of the current position, is the position state variable corresponding to the th moment in the time series. By performing an optimal path search on the joint probability model, map the position state with a probability value exceeding the preset threshold to the coordinates in the geographical coordinate system and output the position coordinates of the suspected termite nest.

[0054] Furthermore, the thermal imaging module obtains the thermal imaging image of the dam surface. Specifically: the thermal imaging module uses a dual-band infrared detector to simultaneously collect the thermal radiation signals in the 3-5μm mid-wave infrared band and the 8-14μm long-wave infrared band. During the collection process, the surface of the dam is scanned point by point through the galvanometer scanning mechanism with a spiral scanning path, and the scanning angle range covers the ±60° field of view directly below the drone; during the scanning process, combine the distance information obtained in real time by the lidar altimetry module to correct the geometric distortion of the thermal imaging image. Establish a background temperature dynamic model through the background suppression algorithm of adaptive Kalman filtering. The formula is: , where is the predicted background temperature at the current moment, is the predicted value at the previous moment, is the adaptive Kalman gain, is the currently collected temperature value, reduce the influence of environmental temperature changes on imaging, and perform pixel-level fusion on the dual-band thermal radiation signals. The formula is: , generate the final thermal imaging image, where and are the mid-wave and long-wave infrared images respectively, , are the corresponding dynamic adjustment coefficients.

[0055] Further, the process of secondary verification of the suspected termite nest position coordinates by combining the thermal imaging image of the dam surface obtained by the thermal imaging module is as follows. The suspected termite nest position coordinates output by the deep learning neural network model are converted into the pixel coordinate system of the thermal imaging image. The mapping relationship between the geographic coordinate system and the image coordinate system is established through the coordinate transformation matrix. Taking the converted pixel coordinates as the center, a local area of pixel size is delimited, and the temperature distribution data within the area is extracted. Through the abnormal temperature area recognition algorithm of density peak clustering, the local density and distance of each pixel point within the area are calculated, and the product of the local density and distance exceeds the set threshold The pixel points are clustered into abnormal temperature areas. If there is such an abnormal temperature clustering area within the delimited local area, it is determined that the suspected position is the termite nest position; if not, the suspected position is excluded, and the result is fed back to the ground control terminal.

[0056] Further, the communication module adopts the ultra-wideband wireless transmission technology of the time-division duplex mode to build a local Internet of Things for the dam. The position information data frame is encapsulated by the Beidou short message protocol and directionally transmitted by the MIMO antenna array after channel coding, realizing the real-time penetration and anti-interference transmission of centimeter-level positioning data.

[0057] This embodiment details the construction of a detection system by carrying multiple modules such as transient electromagnetic by an unmanned aerial vehicle to achieve efficient and accurate positioning of termite nests in dams. The unmanned aerial vehicle can quickly cover a large area, and multi-frequency transmission and multi-sensor collaboration improve the detection efficiency; deep learning and thermal imaging secondary verification accurately identify the characteristics of termite nests and reduce misjudgment; non-contact detection avoids damaging the dam structure and can adapt to complex terrains, and the communication module realizes real-time data transmission, providing reliable technical support for termite prevention and control of dams.

[0058] Based on Embodiment 1, this embodiment details the inertial navigation module, lidar altimetry module, and thermal imaging module of the unmanned aerial vehicle transient electromagnetic detection system, specifically:

[0059] In the method for locating termite nests in dams using a drone-mounted transient electromagnetic system, the inertial navigation module can collect the acceleration and angular velocity data of the drone in three-dimensional space in real time and at high frequencies through built-in accelerometers and gyro inertial sensors. The accelerometer can accurately measure the acceleration of the drone along each axis during flight, thereby calculating the speed and displacement changes of the drone. The gyroscope focuses on detecting the changes in the attitude angles of the drone, including the pitch angle, roll angle, and yaw angle, providing key information for the flight attitude control of the drone. During the dam detection process, due to the complex flight environment, factors such as air flow disturbance and terrain undulation can cause the drone to have attitude offsets. The inertial navigation module can quickly sense these changes and transmit the collected data to the flight control system of the drone. The flight control system adjusts the flight attitude and trajectory of the drone in real time based on this data to ensure that the drone always flies stably along the preset detection route, thereby ensuring that the transient electromagnetic emission and reception devices collect signals at the best position and angle. In addition, the accurate position and attitude information recorded by the inertial navigation module can also be accurately matched with the transient electromagnetic signal acquisition data, making the subsequent data processing and termite nest location results more accurate and reliable, and avoiding detection deviations caused by the position and attitude errors of the drone.

[0060] The lidar altimetry module monitors the distance between the drone and the dam surface in real time and accurately by emitting laser beams towards the dam surface and receiving the reflected laser signals, and calculates the distance between the drone and the target surface based on the flight time of the laser. In the dam detection scenario, the terrain of the dam surface is complex, with slope changes, unevenness, etc. If the flight height of the drone is not properly controlled, it will not only affect the acquisition quality of the transient electromagnetic signals but also pose a safety risk of the drone colliding with the dam. The lidar altimetry module continuously obtains distance data at an extremely high sampling frequency and can quickly respond to terrain changes. When it detects that the terrain height in front is rising, the module quickly transmits the distance change information to the drone control system, and the control system immediately adjusts the flight height of the drone to keep it within the preset safe detection height range; conversely, when the terrain height drops, it can also promptly reduce the flight height to ensure that the distance between the detection device and the dam surface is always in the optimal state, thereby ensuring the stable acquisition of the transient electromagnetic signals. The height data obtained by the lidar altimetry module can also be used for elevation correction of the collected electromagnetic signals. Combining with the position information provided by the inertial navigation module, it can construct a more accurate three-dimensional model of the dam underground structure, providing a richer and more accurate spatial information basis for the location of termite nests.

[0061] The thermal imaging module provides multi-dimensional detection information. Using infrared detection technology, it can capture the infrared radiation emitted by objects on the dam surface and convert it into a visualized thermal imaging image. Since termites generate metabolic heat during the activity in their nests, and the materials and structures of termite nests are different from those of the surrounding soil, the heat conduction and thermal radiation characteristics in the nest area are different from those in the surrounding normal areas, which are manifested as temperature anomaly areas in the thermal imaging image. The thermal imaging module simultaneously collects thermal radiation signals in the 3-5 mid-wave infrared band and 8-14 long-wave infrared band through a dual-band infrared detector. These two bands have their own advantages in sensitivity and penetration ability for objects at different temperatures. Through pixel-level fusion technology, a clearer and more accurate thermal imaging image can be obtained. During detection, the thermal imaging module cooperates with the flight route of the drone to comprehensively scan the dam surface and quickly identify potential temperature anomaly areas. The information of these areas is cross-validated with the suspected termite nest positions obtained by transient electromagnetic detection. When the temperature anomaly areas in the thermal imaging image match the electromagnetic signal anomaly areas, the accuracy and reliability of termite nest positioning can be greatly improved. In addition, the thermal imaging module also has the characteristic of being not restricted by lighting conditions. Whether it is day or night, it can continuously and stably monitor the temperature distribution on the dam surface, providing a strong guarantee for the all-weather detection of termite nests on the dam.

[0062] This embodiment details that the inertial navigation module of the drone transient electromagnetic detection system in this application can calibrate the attitude and trajectory of the drone in real time to ensure the accuracy of the detection route; the lidar altimetry module dynamically adjusts the flight height to ensure that the detection device is at the best working distance; the thermal imaging module uses infrared radiation to capture temperature anomalies and combines electromagnetic signal cross-validation. The three work together to greatly improve the efficiency, accuracy and reliability of termite nest positioning on the dam.

[0063] Based on Embodiment 1, this embodiment details the secondary verification of the suspected termite nest position coordinates using the thermal imaging image of the dam surface obtained by the thermal imaging module. Specifically:

[0064] Convert the suspected termite nest position coordinates output by the deep learning neural network model into the pixel coordinates of the thermal imaging image, and establish the connection between the two through the coordinate transformation matrix The formula is , represents the geographical coordinates, which are obtained by the high-precision positioning system carried by the drone in combination with the inertial navigation module and can accurately determine the actual physical position on the dam; is the converted pixel coordinate, corresponding to the specific position point in the thermal imaging image, and the coordinate transformation matrix It is obtained by pre-calibrating the thermal imaging module, taking into account factors such as the flight altitude, attitude angle of the drone, and distortion parameters of the thermal imaging lens. During actual operation, the attitude of the drone changes continuously, and the distance information monitored in real time by the lidar altimetry module is also fed back to the coordinate conversion process, for to perform dynamic correction to ensure the accuracy of the conversion, in order to delimit Taking the local area with a pixel size as an example, if precise coordinate conversion is not performed, resulting in the misalignment of the extracted area with the actual suspected termite nest location, thus affecting the subsequent verification results. Only based on precise coordinate mapping can it be ensured that the local area extracted from the thermal imaging image actually covers the suspected location determined by the deep learning model;

[0065] After completing the coordinate conversion and delimiting the local area, an abnormal temperature region recognition algorithm based on density peak clustering is used to process the temperature distribution data in this area, and calculate the local density of each pixel point in the area and the distance , where, is the Euclidean distance of the temperature between pixel point and , which reflects the magnitude of the temperature difference between two points, As the cut-off distance, it is a threshold preset according to a large amount of experimental data and the temperature distribution characteristics of the dam surface, used to define the degree of proximity of the distance between pixel points; is the indicator function. When is less than 0, takes the value of 1, otherwise it is 0. Through the indicator function, the number of pixel points with a distance less than from pixel point can be counted, and then the local density is obtained. The higher the local density, the denser the pixel points with similar temperatures around this point; represents the minimum distance from pixel point to the pixel points with a larger local density than it. This parameter is used to measure the relative isolation degree of the pixel point in the temperature distribution. Pixel points with the product of the local density and the distance exceeding the threshold are clustered into abnormal temperature regions, and the threshold It is trained through machine learning algorithms by combining historical thermal imaging data of termite nests in dikes and data of normal areas, and can effectively distinguish temperature anomalies caused by termite activities from temperature fluctuations caused by environmental factors. Affected by environmental factors such as sunlight and wind direction, there will be certain temperature changes on the surface of the dike, but these changes are usually continuous and uniform. Due to the biological activities and special structure inside the nest, the temperature distribution in the termite nest area will show the characteristics of local aggregation and obvious difference from the surrounding area. Through the density peak clustering algorithm, this abnormal temperature distribution can be accurately captured. If there is a qualified abnormal temperature clustering area in the designated local area, then the suspected location is determined as the location of the termite nest; if not, the suspected location is excluded, thus effectively verifying the output result of the deep learning model and improving the reliability and accuracy of termite nest positioning.

[0066] In this embodiment, it is described in detail that through coordinate precise mapping and density peak clustering algorithm, the suspected termite nest location output by deep learning is associated with the thermal imaging image. The coordinate transformation matrix is dynamically corrected to ensure accurate area correspondence. The clustering algorithm combines the temperature distribution characteristics to accurately identify anomalies, effectively eliminates environmental interference, realizes double verification, and significantly improves the accuracy and reliability of termite nest positioning.

[0067] Based on Embodiment 1, this embodiment details the specific implementation process of this application, specifically:

[0068] In terms of detection efficiency, the drone in this application has a shorter detection duration compared with ground penetrating radar and resistivity meters, and the detection method is more convenient, shortening the detection cycle, and having significant advantages in the large-area dike detection scenario.

[0069] As Figures 3 - 4 shown, positioning accuracy is the core index to measure the termite nest detection technology of dikes. The technology in this application uses multi-frequency transient electromagnetic signal acquisition, wavelet transform denoising and deep learning neural network recognition, with a positioning accuracy rate as high as 94.29% and a missed judgment rate of only 5.71%; while the accuracy rate of ground penetrating radar is 59.52% and the missed judgment rate is 28.57%; the accuracy rate of resistivity meter is 57.89% and the missed judgment rate is 39.47%; the technology in this application has 8 more accurately identified locations than ground penetrating radar and 11 more than resistivity meters, and the number of misjudgments is 10 and 8 less than the two respectively. The secondary verification mechanism of the thermal imaging module further improves the reliability. For the 35 suspected locations output by the deep learning model, the verification success rate reaches 80%, effectively reducing the risk of misjudgment.

[0070] As shown in Figure 5, the improved deep learning neural network of this application adopts a dual-channel attention convolutional neural network structure. During the training process, compared with the traditional convolutional neural network, the convergence speed is increased by 37%. Taking the training of 100 epochs as an example, it takes about 0.23 hours for the loss value of the traditional convolutional neural network to drop to 0.35, while the improved model of this application only takes 0.18 hours. In terms of the performance on the test set, the model of this application has better ability to extract the features of complex electromagnetic signals, with an accuracy rate of 92.34%, while the accuracy rate of the traditional convolutional neural network is only 78.65%. Looking at the existing ground penetrating radar technology, in the area with a soil water content of 35%, the detection depth drops suddenly from 3m to 1.2m, and the resolution drops by 40%; in the electromagnetic interference environment, the data misjudgment rate of the resistivity meter is as high as 45%. Both are difficult to complete the task of detecting termite nests in dams stably and accurately. However, the improved model of this application can still maintain high-efficiency feature extraction and positioning ability in a complex electromagnetic environment, with significant performance advantages.

[0071] In terms of the anti-environmental interference performance, the technology of this application also performs excellently. In an environment with a wind speed of 5.6m / s, the inertial navigation module and the lidar altimetry module work together to ensure that the flight altitude fluctuation of the drone is controlled within ±0.3m, and the signal acquisition stability reaches 98.7%. While for the ground penetrating radar in the same environment, the signal distortion rate is as high as 23%, and the data availability rate of the resistivity meter drops to 65%. Generally speaking, whether it is the detection efficiency, positioning accuracy, or data processing ability and environmental adaptability, the various values of the technology of this application are much better than the existing technology, fully demonstrating its innovation and practicality.

[0072] This embodiment details that the technology of this application, relying on multi-module cooperation and advanced algorithms, demonstrates excellent performance in the detection of termite nests in dams. Its detection efficiency, positioning accuracy rate, and missed judgment rate far exceed the existing technology. It has efficient data processing and strong anti-interference ability, and still maintains high stability in a complex environment, being comprehensively superior to traditional detection means.

[0073] The above are only the preferred embodiments of the present invention, and it does not limit the protection scope of the present invention accordingly. For those skilled in the art, the present invention can have various changes and modifications; all changes, modifications, substitutions, integrations, and parameter changes made to these embodiments by means of conventional substitutions or capable of achieving the same functions without departing from the principle and spirit of the present invention fall within the protection scope of the present invention.

Claims

1. A method for locating termite nests in dikes based on transient electromagnetic carried by drones, characterized in that, Including: Construct a drone transient electromagnetic detection system; Preprocess the received secondary induced electromagnetic field signals, calculate the weights of signals at different frequencies according to the response differences of multi-frequency signals, and fuse to obtain a comprehensive electromagnetic signal; Input the comprehensive electromagnetic signal into a pre-trained deep learning neural network model, identify the characteristic signals related to termite nests in the comprehensive electromagnetic signal, and output the position coordinates of suspected termite nests; Combine the thermal imaging image of the dam surface obtained by the thermal imaging module to conduct a secondary verification on the position coordinates of suspected termite nests output by the deep learning neural network model. If there is an abnormal temperature distribution area at the corresponding position in the thermal imaging image, determine that position as the termite nest position, and transmit the termite nest position information to the ground control terminal in real time through the communication module carried by the drone.

2. The method for locating termite nests in dikes based on transient electromagnetic detection carried by an unmanned aerial vehicle according to claim 1, wherein The drone transient electromagnetic detection system adopts a multi-frequency time-division transmission technology to transmit at least three different frequencies of pulsed current signals; multiple independent transmit waveform generation modules are arranged in the transient electromagnetic detection system, each module corresponding to a pulsed current signal of a different frequency. The programmable logic controller controls the timing switching circuit to sequentially activate each transmit waveform generation module at a preset time interval; within each transmit cycle, first transmit a low-frequency pulsed current signal for detecting the deep area of the dam, transmit a medium-frequency pulsed current signal after an interval of 5 - 10 milliseconds for detecting the middle area, and then transmit a high-frequency pulsed current signal after an interval of 3 - 5 milliseconds for detecting the shallow area. The transmit duration and current intensity of each frequency signal are adaptively adjusted according to the geological conditions of the dam and historical detection data to obtain effective excitation and detection in different depth areas.

3. The method for locating termite nests in dikes based on transient electromagnetic carried by an unmanned aerial vehicle according to claim 2, wherein The drone transient electromagnetic detection system further includes an inertial navigation module, a lidar altimetry module, and a thermal imaging module; the inertial navigation module is used to obtain the attitude and position information of the drone in real time; the lidar altimetry module is used to monitor the distance between the drone and the dam surface in real time; the thermal imaging module is used to assist in identifying abnormal heat source areas on the dam surface.

4. The method for locating termite nests in dikes based on transient electromagnetic field carried by unmanned aerial vehicle according to claim 1, wherein The preprocessing of the secondary induced electromagnetic field signal is specifically as follows: the original signal collected is processed by the wavelet transform denoising algorithm to be decomposed into high-frequency components of different scales and low-frequency components , and the formula is: , where is the number of wavelet decomposition layers. An adaptive threshold is set for the high-frequency components . The improved soft threshold function is used to denoise each layer of high-frequency components to generate the denoised high-frequency components , and the formula is: ， Among them, is the sign function, and the denoised high-frequency component and the low-frequency component are reconstructed through inverse wavelet transform to obtain the denoised signal .

5. The method for locating termite nests in dikes based on transient electromagnetic carried by drones according to claim 1 or 4, characterized in that, Based on the denoised signal, calculate the weights of signals with different frequencies according to the response differences of multi-frequency signals , and the formula is: , where is the standard deviation of the -th frequency signal, is the number of frequency signals; perform weighted summation on each frequency signal through to obtain the comprehensive electromagnetic signal , and the formula is: , where is the denoised signal of the -th frequency.

6. The method for locating termite nests in dikes based on transient electromagnetic carried by drones according to claim 1, characterized in that, The pre-trained deep learning neural network model adopts a dual-channel attention convolutional neural network structure, including parallel frequency feature channels and time series channels; The frequency feature channel adopts a multi-layer grouped convolution module, and the convolution kernel size and dilation rate of each group of convolutional layers are adaptively adjusted according to the characteristics of multi-frequency signals for extracting the characteristics of electromagnetic signals at different frequencies; The time series channel adopts a structure combining a long short-term memory network and a self-attention mechanism to capture the sequence characteristics of electromagnetic signals changing over time; Set a cross-channel attention module after two channels. By calculating the mutual information between the frequency features and the time series features, a channel attention weight matrix is generated. , the formula is , where is the frequency feature vector, is the time series feature vector, and are linear transformation functions respectively; After concatenating the weighted features of the two channels, input them into the global average pooling layer and the fully connected layer, and use the focal loss function for training, where and are hyperparameters, is the probability of termite mounds predicted by the model; The training data includes the electromagnetic signal data of the dikes under different geological conditions and different termite damage levels and the corresponding annotation information of the actual positions of the termite mounds, enhancing the recognition ability of the characteristic signals of the termite mounds on the dikes.

7. The method for locating termite nests in dikes based on transient electromagnetic carried by unmanned aerial vehicle according to claim 1 or 6, characterized in that, The process of identifying the characteristic signals related to termite nests in the comprehensive electromagnetic signals through the deep learning neural network model and outputting the suspected termite nest location coordinates is as follows: The comprehensive electromagnetic signals are vectorized according to the time series and frequency dimensions and converted into a three-dimensional data tensor , where is the number of time sampling points, is the number of frequency channels, is the number of data channels; the data tensor is input into the deep learning neural network model. The local electromagnetic features at different frequencies are extracted by the grouped convolutional layer of the frequency feature channel, and the LSTM-self-attention module of the time series channel captures the variation law of the electromagnetic signals over time. Then, the cross-channel attention module assigns weights and fuses the features of the two channels; at the model output layer, through the position decoding mechanism of the conditional random field, a joint probability model of the characteristic signals and spatial positions is constructed. The formula is: , where is the position state sequence, is the normalization factor, is the transition potential function between adjacent positions, is the characteristic potential function of the current position, is the position state variable corresponding to the th moment in the time series. By performing an optimal path search on the joint probability model, the position states with probability values exceeding the preset threshold are mapped to the coordinates in the geographical coordinate system, and the suspected termite nest location coordinates are output.

8. The method for locating termite nests in dikes based on transient electromagnetic carried by an unmanned aerial vehicle according to claim 1, characterized in that The thermal imaging module obtains the thermal imaging image of the dam surface. Specifically: The thermal imaging module uses a dual-band infrared detector to simultaneously collect thermal radiation signals in the 3-5μm mid-wave infrared band and the 8-14μm long-wave infrared band. During the collection process, a galvanometer scanning mechanism is used to perform point-by-point scanning of the dam surface along a spiral scanning path, and the scanning angle range covers the ±60° field of view directly below the drone. During the scanning process, combined with the distance information obtained in real time by the lidar altimetry module, geometric distortion correction is performed on the thermal imaging image. Through the background suppression algorithm of adaptive Kalman filtering, a dynamic background temperature model is established, and the formula is: , where is the predicted background temperature at the current moment, is the predicted value at the previous moment, is the adaptive Kalman gain, is the temperature value collected currently, to reduce the influence of ambient temperature changes on imaging, the dual-band thermal radiation signals are pixel-level fused, and the formula is: , generating the final thermal imaging image, where and are the mid-wave and long-wave infrared images respectively, , are the corresponding dynamic adjustment coefficients.

9. The method for locating termite nests in dikes based on transient electromagnetic detection carried by an unmanned aerial vehicle according to claim 1 or 8, characterized in that, The process of using the thermal imaging image of the dam surface obtained by the combined thermal imaging module to perform secondary verification on the suspected termite nest location coordinates is as follows. The suspected termite nest location coordinates output by the deep learning neural network model are converted into the pixel coordinate system of the thermal imaging image. The mapping relationship between the geographic coordinate system and the image coordinate system is established through a coordinate transformation matrix. Taking the converted pixel coordinates as the center, a local area of pixel size is delimited, and the temperature distribution data within the area is extracted. Through the abnormal temperature area recognition algorithm of density peak clustering, the local density and distance of each pixel point within the area are calculated, and the pixel points whose product of the local density and distance exceeds the set threshold are clustered into abnormal temperature areas. If there is such an abnormal temperature clustering area within the delimited local area, it is determined that the suspected location is the termite nest location; if not, the suspected location is excluded, and the result is fed back to the ground control terminal.

10. The method for locating termite nests in dikes based on transient electromagnetic carried by an unmanned aerial vehicle according to claim 1, characterized in that, The communication module adopts an ultra-wideband wireless transmission technology in a time-division duplex mode to construct a local Internet of Things for the dam, encapsulate the position information data frame through the Beidou short message protocol, and perform directional transmission by the MIMO antenna array after channel coding to achieve real-time penetration and anti-interference transmission of centimeter-level positioning data.

Images