Dam termite nest accurate positioning method and device based on artificial intelligence and medium

Abstract

The invention relates to the field of dam termite detection, and discloses a dam termite nest accurate positioning method and device based on artificial intelligence and a medium, and the system comprises a ground listening detector and data acquisition module and a neural network processing module; the earth hearing detector and the data acquisition module are used for acquiring and processing termite nest sound wave signals; and the neural network processing module accurately positions the termite nest position according to the termite nest sound wave signal. According to the method, the signal intensity information extracted by the deep learning model is fully utilized, and the precise estimation of the termite nest position is realized in combination with the physical model and the optimization algorithm. The signal intensity output by the deep learning model is used as the input of the physical model, so that challenges caused by noise interference and signal complexity are solved, and the accuracy and reliability of a positioning result are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of dam termite detection, and in particular to a dam termite nest precise positioning method, device and medium based on artificial intelligence. BACKGROUND

[0002] Dam termites are a type of termite that inhabits water dams, embankments, bridges and other water conservancy projects. They usually build nests in the soil and build nests by eating wood and plant fibers, which can cause serious damage to the structure of water conservancy projects. In addition, the activities of dam termites can cause soil loosening and erosion, affecting the overall stability and safety of the project.

[0003] In the field of dam termite detection, existing acoustic positioning technology faces many challenges. Because soil is a good acoustic conductor, the data collected by the detector often contains a large amount of noise, so in the process of removing noise, part of the effective signal will inevitably be lost. In addition, the acoustic signal of the termite nest is extremely complex and not composed of a single frequency, making it difficult for traditional signal processing methods to accurately distinguish. Combined with the limitations of the sensitivity of the detector, the difference in signal strength of the termite nest sound source received by each detector is small, further increasing the difficulty of positioning. SUMMARY

[0004] The purpose of the present application is to provide a dam termite nest precise positioning method, device and medium based on artificial intelligence, which solves the technical problem that the current traditional method cannot accurately position the dam termite nest.

[0005] Specifically, the present application provides a dam termite nest precise positioning system based on artificial intelligence, which comprises the following: Ground sound detector, data acquisition module and neural network processing module; The ground sound detector and data acquisition module are used to collect and process termite nest acoustic signals; The neural network processing module accurately locates the position of the termite nest according to the termite nest acoustic signal.

[0006] A dam termite nest precise positioning method based on artificial intelligence, the method comprising the following steps: S1, collecting and processing termite nest acoustic signals by a ground sound detector and data acquisition module; S2, pre-training the neural network processing module using a loss function to obtain a trained neural network processing module; S3, inputting the termite nest acoustic signal into the trained neural network processing module to obtain the position of the termite nest.

[0007] A storage medium stores instructions and data for implementing a dam termite nest positioning method.

[0008] A dam termite nest positioning device comprises a processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement a dam termite nest positioning method.

[0009] The present application provides the beneficial effect of: making full use of the signal strength information extracted by the deep learning model, combining the physical model and the optimization algorithm, and realizing the accurate estimation of the position of the termite nest. By taking the signal strength output by the deep learning model as the input of the physical model, the challenge brought by noise interference and signal complexity is solved, and the accuracy and reliability of the positioning result are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 is a schematic diagram of the system structure of the present application; Figure 2 is a schematic diagram of the neural network processing module structure of the present application; Figure 3 is a schematic diagram of the termite nest positioning result of the present application; Figure 4 is a schematic diagram of the hardware device working of the embodiment of the present application. DETAILED DESCRIPTION

[0011] In order to make the purpose, technical scheme and advantages of the present application clearer, the embodiments of the present application will be further described below with reference to the drawings.

[0012] Before formally describing the present application, the scheme of the present application will be described generally first, for easy understanding.

[0013] Please refer to Figure 1 The present application provides a dam termite nest precise positioning system based on artificial intelligence, which comprises: a ground listening detector, a data acquisition module and a neural network processing module; The ground listening detector and the data acquisition module are used for collecting and processing termite nest sound wave signals. It should be noted that the ground listening detector and the data acquisition module comprise a high-sensitivity ground listening detector array and a low-noise photoelectric signal detection and processing sub-module. The high-sensitivity ground listening detector array is used for high-sensitivity detection of termite nest signals; the low-noise photoelectric signal detection and processing sub-module is used for reducing noise in termite nest signals.

[0014] Specifically, the high-sensitivity ground listening detector array adopts an optical fiber interferometer to realize high-sensitivity detection of termite nest signals. The low-noise photoelectric signal monitoring processing sub-module adopts a 3x3 optical fiber coupler demodulation method, effectively reduces the noise of the detector, ensures the stability and accuracy of the signal, and meets the needs of weak signal detection in the dam environment. The neural network processing module accurately locates the termite nest position according to the termite nest sound wave signal.

[0015] Please refer to Figure 2 It should be noted that the neural network processing module adopts an optimized SqueezeNet architecture as the basic network.

[0016] Specifically, in order to adapt to single-channel audio input, the first layer of convolutional layer of the model is modified to accept single-channel input, and the convolution kernel size is 3x3 and the step is 2. The model retains the efficient feature extraction capability of SqueezeNet, and uses an adaptive average pooling layer (AdaptiveAvgPool2d) to reduce the feature map to a spatial dimension of 1x1.

[0017] In the classifier part, a fully connected layer is designed to map the extracted feature vector to an output for judging the intensity of termite activity. To prevent overfitting, a Dropout layer is added to the model to randomly shield the output of part of the neurons. During the model training process, appropriate loss function and optimization algorithm are used, and the model is trained on a large number of labeled termite audio data sets. By adjusting the learning rate and batch size and other hyperparameters, the generalization ability of the model is further improved.

[0018] A dam termite nest precise positioning method based on artificial intelligence, based on the positioning system, comprising the following steps: S1, collecting and processing the termite nest sound wave signal through the ground sound detector and the data acquisition module; In actual application process, first, the ground sound detector array is used to collect the vibration signal inside the dam. Since the weak sound wave signal generated by termite activity is complex and easy to be disturbed by noise, the low-noise photoelectric signal monitoring processing sub-module is used to effectively reduce the noise of the detector, ensure the stability and accuracy of the signal, and meet the needs of weak signal detection in the dam environment. S2, pre-training the neural network processing module using a loss function to obtain a trained neural network processing module; In the model training process, a customized loss function is designed to improve the performance of the model in the ranking task.

[0019]

[0020]

[0021] wherein is an audio signal, is the result of the model calculation, is the result of the model calculation after normalization, is the distance from the audio signal collection point to the termite nest, is the loss function. The loss function is based on the idea of learning ranking, which processes the model output and the real target. Specifically, first, apply the Softmax function to the output of the model to get the predicted probability distribution. Then, take the reciprocal of the square of the real target value to get the weight of the target, and normalize the weights to ensure that their sum is 1. Finally, calculate the cross-entropy loss between the predicted probability distribution and the target weight distribution.

[0022] This method can effectively emphasize the signals of termite nests that are close in distance, and better guide the training of the model.

[0023] Secondly, in the application process, a specific function model is designed to establish a connection between the model output and the actual physical quantity. An inverse square function model is used, which means that the signal strength is inversely proportional to the square of the distance. By fitting the predicted signal strength and actual distance data, the model parameters are estimated, and the precise positioning of the termite nest location is achieved. In this process, the mean and standard deviation of the model prediction values at different distances are calculated, and the curve fitting method is used to evaluate and optimize the prediction performance of the model.

[0024] Specifically, the signal strength is inversely proportional to the square of the distance from the termite nest to the detector, following the inverse square law in physics. The relationship model between signal strength I and distance r is established:

[0025] where, A is the source intensity parameter to be estimated, r is the distance between the nest and the detector.

[0026] In order to solve the position coordinates and source intensity A of the nest, a function is constructed, which represents the error between the model-predicted signal strength and the signal strength output by the deep learning model:

[0027] The L-BFGS-B optimization algorithm is used to minimize the loss function to obtain the position coordinates and source intensity parameters of the nest. To improve the stability of the optimization and avoid falling into local minimum, multiple random initializations are performed, and the initial parameters are randomly generated in each iteration, with a total of multiple iterations. The initial parameters include three spatial coordinates and source intensity.

[0028] S3, input the termite nest sound wave signal to the trained neural network processing module, and perform position fitting to obtain the position of the termite nest.

[0029] As an embodiment, the ground listening detector array in the present application is composed of a plurality of high-sensitivity ground listening detectors arranged at different positions of the dam for collecting weak sound wave signals emitted by termite nests.

[0030] Each ground listening detector adopts a fiber interferometer structure to improve the detection sensitivity of termite activity signals.

[0031] The detector is connected to a low-noise photoelectric signal monitoring and processing sub-module, and a 3x3 fiber coupler demodulation method is used to effectively reduce the noise of the detector itself and ensure the stability and accuracy of the signal.

[0032] Subsequently, the noise-reduced signal is input to the neural network processing module. A modified SqueezeNet is used as the base network, the first convolutional layer is adapted to single-channel audio input, the convolution kernel size is 3x3, and the step size is 2. Adaptive average pooling layers and Dropout layers are added to the model to prevent overfitting and improve feature extraction capability.

[0033] During model training, a large amount of labeled termite audio data is used, and a customized loss function based on the learning ranking idea is used. The loss function calculates the cross-entropy loss between the predicted probability distribution and the target weight distribution, emphasizing the signals of termite nests with close distances, and better guiding the training of the model. The optimization algorithm selects an appropriate optimizer, adjusts the learning rate and batch size, and improves the generalization ability and accuracy of the model.

[0034] In practical application, the ground listening detector array continuously monitors the vibration signals inside the dam. The collected signals are preprocessed and input to the deep learning model, and the model outputs signal intensity values representing termite activity intensity. These signal intensity values reflect the influence of termite nests on each detector.

[0035] Assuming that the signal intensity is inversely proportional to the square of the distance from the termite nest to the detector, following the inverse square law in physics, a relationship model between signal intensity I and distance r is established. The L-BFGS-B optimization algorithm is used to minimize the loss function, and the position coordinates and source intensity parameters of the nest are iteratively updated. To improve the stability of the optimization and avoid falling into local minimum, multiple random initializations are performed, and the initial parameters, including spatial coordinates, are randomly generated each iteration. Through the above method, the system can achieve high-precision positioning of termite nests, effectively solving the challenges brought by noise interference and signal complexity, and improving the accuracy and reliability of the positioning results.

[0036] The system described in this embodiment is used in the following way: first, arrange the array of geophone detectors in the dam area, ensuring that the detectors are in close contact with the dam structure. Start the data acquisition module to collect the acoustic wave signals transmitted by the detectors in real time. The pre-trained neural network processing module analyzes the signals and outputs the intensity information of termite activity. According to the relationship between signal intensity and distance, the specific location of the termite nest is calculated using an optimization algorithm.

[0037] Please refer to Figure 3 , Figure 3 is a schematic diagram of the termite nest positioning result of the present invention.

[0038] The system described in this embodiment is not limited to a specific type of dam structure or a specific species of termites. By adjusting the arrangement of the detectors, optimizing the parameters of the deep learning model, and other means, it can adapt to the detection needs of termite nests in different environments.

[0039] Please refer to Figure 4 , Figure 4 is a hardware device working schematic diagram of the embodiment of the present invention, which specifically includes: an artificial intelligence-based dam termite nest precise positioning device 401, a processor 402, and a storage medium 403.

[0040] An artificial intelligence-based dam termite nest precise positioning device 401: the artificial intelligence-based dam termite nest precise positioning device 401 realizes the artificial intelligence-based dam termite nest precise positioning method.

[0041] Processor 402: the processor 402 loads and executes the instructions and data in the storage medium 403 to realize the artificial intelligence-based dam termite nest precise positioning method.

[0042] Storage medium 403: the storage medium 403 stores instructions and data; the storage medium 403 is used to realize the artificial intelligence-based dam termite nest precise positioning method.

[0043] The beneficial effects of the present invention are: fully utilizing the signal intensity information extracted by the deep learning model, combining the physical model and the optimization algorithm, realizing the accurate estimation of the position of the termite nest. By taking the signal intensity output by the deep learning model as the input of the physical model, the challenges brought by noise interference and signal complexity are solved, and the accuracy and reliability of the positioning result are improved.

[0044] The above is only the preferred embodiment of the present invention, and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. An artificial intelligence-based dam termite nest precision positioning system, characterized in that: include: Ground-sensing detector, data acquisition module, and neural network processing module; The ground-listening detector and data acquisition module are used to collect and process acoustic signals from termite nests; The neural network processing module accurately locates the termite nest based on the acoustic signals from the termite nest.

2. The AI-based embankment termite nest precise positioning system of claim 1, wherein: The ground-sensing detector and data acquisition module includes: a high-sensitivity ground-sensing detector array and a low-noise photoelectric signal detection and processing submodule; the high-sensitivity ground-sensing detector array is used to achieve high-sensitivity detection of termite nest signals; the low-noise photoelectric signal detection and processing submodule is used to reduce noise in termite nest signals.

3. The AI-based embankment termite nest precise positioning system of claim 2, wherein: The high-sensitivity ground hearing detector array uses a fiber optic interferometer.

4. The AI-based embankment termite nest precise positioning system of claim 2, wherein: The low-noise photoelectric signal detection and processing submodule uses a 3×3 fiber optic coupler demodulation method to reduce noise in the termite nest signal.

5. The AI-based embankment termite nest precise positioning system of claim 1, wherein: The neural network processing module uses an optimized SqueezeNet architecture as its base network.

6. The AI-based embankment termite nest precise positioning system of claim 5, wherein: The optimized SqueezeNet architecture specifically refers to: The first convolutional layer of the model was modified to accept single-channel input, and an adaptive average pooling layer was used to shrink the feature map to a 1×1 spatial dimension. The classifier part is designed with a fully connected layer to determine the intensity of termite activity.

7. A method for precise positioning of embankment termite nest based on artificial intelligence, which is implemented based on the positioning system according to any one of claims 1-6. The method includes the following steps: S1. Acquire and process acoustic signals from termite nests using a ground-sensing detector and a data acquisition module; S2. Train the neural network processing module using the loss function to obtain the trained neural network processing module; S3. Input the acoustic signal of the termite nest into the trained neural network processing module and perform position fitting to obtain the location of the termite nest.

8. The method of claim 7, wherein the method is based on artificial intelligence. In step S2, the trained model predicts the signal strength information of the sound wave signal from the termite nest to the detection point based on the detected sound wave signal. The loss function is as follows: wherein is an audio signal, is a calculation result of a model, is signal strength information corresponding to the signal transmitted by the termite nest to the detection point, is the distance from the audio signal collection point to the ant nest, is a loss function, is the number of samples for training; Step S3 uses the model trained in step S2 to obtain the corresponding acoustic signal intensity information of the termite nest based on the detected acoustic signals, and substitutes it into the function: wherein is the spatial position of the detection point; is the actual termite nest position; A is the termite nest sound wave signal strength; N is the sample number; is the sound wave signal corresponding to the termite nest sound wave signal strength information. By minimizing the function the spatial position of the termite nest is obtained.

9. A storage medium characterized by: The storage medium stores instructions and data to implement the artificial intelligence-based method for accurately locating termite nests in dikes as described in any one of claims 7-8.

10. An artificial intelligence-based dam termite nest precise positioning device, characterized in that: include: A processor and a storage medium; the processor loads and executes instructions and data in the storage medium to implement the artificial intelligence-based method for accurately locating termite nests in dams as described in any one of claims 7-8.

Images