Intelligent termite detection system based on MIMO radar technology and deep learning

Abstract

This invention discloses an intelligent termite detection system based on MIMO radar technology and deep learning, belonging to the field of target recognition technology. The invention includes a housing device and a mobile terminal. The housing device is deployed near the target and uses microwave radar to detect the target, acquiring radar echo signals from inside the target and wirelessly transmitting the corresponding radar image and radar energy to the mobile terminal. The mobile terminal identifies termites based on the deployed detection model and determines the termite activity area based on radar energy changes. This invention, by combining MIMO radar technology, deep learning algorithms, and radar energy change detection, significantly improves the accuracy of termite activity detection and offers strong real-time performance. The mobile terminal's detection program is highly adaptable and user-friendly. This invention not only facilitates termite detection but also provides convenience for subsequent processing. Users can view the detection results on the mobile terminal and take corresponding prevention and control measures based on the results.

Description

Technical Field

[0001] This invention relates to the field of target recognition technology, and in particular to an intelligent termite detection system based on MIMO radar technology and deep learning. Background Technology

[0002] Termites are social insects that typically live underground or inside buildings, forming highly organized colonies. They feed on wood and cellulose, and can damage building structures, wooden furniture, and other cellulose-containing materials. Due to their stealth and efficient reproduction, termite activity is often difficult to detect early, leading to serious economic losses and safety hazards.

[0003] The damage caused by termites is mainly manifested in the following aspects:

[0004] Building damage: Termites erode wooden structures, weakening the stability of buildings and potentially causing them to collapse in severe cases.

[0005] Economic losses: Termites are highly destructive, and repairing damaged buildings and furniture requires a significant financial investment.

[0006] Safety hazards: Buildings and furniture infested with termites pose safety hazards and may threaten the personal safety of residents.

[0007] Environmental impact: Termite activity alters soil structure, affects plant growth, and consequently impacts the ecological environment.

[0008] Current termite detection methods suffer from significant shortcomings in terms of location accuracy and operational efficiency, particularly for termite activity hidden within walls. Traditional detection methods rely on human experience, making it difficult to accurately detect concealed areas. In recent years, MIMO radar technology, with its superior penetration capabilities and high resolution, has provided a new technological approach for termite detection. Combining it with deep learning algorithms can further improve detection accuracy and efficiency.

[0009] Existing termite detection systems mostly employ manual or semi-automatic methods, which have limitations in precise location and real-time monitoring, failing to meet practical needs. Because radar signals have penetrating power and can detect objects hidden inside walls, combining them with deep learning technology can achieve accurate identification and monitoring of termite activity. However, existing solutions still have many technical shortcomings in practical applications, such as complex data processing, poor real-time performance, and insufficient adaptability to mobile platforms. Summary of the Invention

[0010] The purpose of this invention is to provide an intelligent termite detection system based on MIMO radar technology and deep learning, which combines microwave radar and deep learning models to achieve real-time monitoring and accurate identification of termite activity.

[0011] The technical solution adopted in this invention is as follows:

[0012] A smart termite detection system based on MIMO radar technology and deep learning, which includes a box-type device and a mobile terminal.

[0013] The enclosure equipment is deployed near the target to detect it using microwave radar and to obtain radar echo signals from inside the target.

[0014] The enclosure includes an outer enclosure, an inner enclosure, a microwave radar device, a Raspberry Pi data processing unit, a power module, and external interfaces. The microwave radar device is fixed to the bottom of the outer enclosure with its radar scanning direction downward. The Raspberry Pi data processing unit, which is connected to the microwave radar device, is located on the top of the outer enclosure. The power module is used to power the microwave radar device and the Raspberry Pi data processing unit. The outer enclosure of the microwave radar device is also equipped with external interfaces, which serve as power interfaces and data transmission interfaces.

[0015] The Raspberry Pi data processing unit is used to set the operating parameters of the microwave radar equipment. The microwave radar equipment is used to collect radar echo signals inside the target and send them to the Raspberry Pi data processing unit. The Raspberry Pi data processing unit is used to filter noise from the radar echo signals, then perform imaging processing to obtain the corresponding two-dimensional image data, and perform image enhancement processing to obtain the data to be detected. The data to be detected is then transmitted to the mobile terminal via a wireless network. Based on the preset division of the image region generated by the echo signal, the Raspberry Pi data processing unit calculates the radar energy of each region based on the received echo signal and transmits it to the mobile terminal via a wireless network.

[0016] The mobile terminal is equipped with a termite detection application based on the MobileNetV2 model, which is used to perform target detection on the received data to be detected and obtain preliminary detection results; and based on the monitoring of changes in radar echo energy, confirm the activity area of ​​termites in the preliminary detection results, generate the final termite detection results, and display them on the mobile terminal.

[0017] This invention enables real-time monitoring and visualization of termite activity of the target based on a box-shaped device deployed near the target (such as a cement wall, wooden wall, or other wall-shaped target), a remote mobile terminal, and a pre-installed user interaction program and a MobileNetV2 deep learning model.

[0018] Furthermore, the user interface of the termite detection application on mobile devices includes:

[0019] Start / Stop Function: This function enables the start and stop of data saving. A "Start / Stop" option can be set at the top of the user interface on the mobile terminal so that users can start or stop the visualization process at any time, avoiding the generation of redundant data.

[0020] Real-time monitoring display area: Used to display real-time video or images transmitted by microwave radar equipment, and to mark the confidence level of the corresponding area when termite activity is detected;

[0021] Data saving function: A "Save" option can be set at the top of the user interface, allowing users to save relevant data after the detection is completed for subsequent analysis.

[0022] Furthermore, the termite detection application's user interface on mobile devices also includes a historical detection record query function, allowing users to view and analyze past detection data, which supports filtering and review based on time criteria.

[0023] Furthermore, noise filtering employs a high-pass filter to remove background noise signals that do not exceed a specified frequency.

[0024] Furthermore, the Raspberry Pi data processing unit performs image enhancement processing using histogram equalization.

[0025] Furthermore, the MobileNetV2 model sequentially includes an input layer, a start convolutional layer, an inverted residual block sequence, a convolutional layer before the end, a local average pooling layer, and a fully connected layer based on the Softmax activation function, which is used to output the termite detection confidence of each image region of the data to be detected, and obtain preliminary detection results.

[0026] Furthermore, the MobileNetV2 model is based on a channel attention mechanism.

[0027] Furthermore, the termite dataset required for training the MobileNetV2 model is constructed as follows:

[0028] Obtain wood samples with diverse shapes and forms that have been eaten by termites;

[0029] In the constructed experimental environment, a wall material of uniform thickness was placed, and two areas were set up behind the wall material. One area was placed with normal wood blocks, and the other area was placed with the wood samples that had been eaten by insects (hereinafter referred to as insect samples).

[0030] The two areas were separated by a partition, and a certain number of termites, such as subterranean termites, were added to the side where the wood sample that had been eaten by insects was placed. The areas were marked on the wall material of uniform thickness to correspond to the image division of the two-dimensional image obtained by radar echo signal imaging.

[0031] The box-shaped device of the present invention is deployed in front of a wall material with uniform thickness to collect the termite dataset required for training the MobileNetV2 model.

[0032] Among them, the wall material with uniform thickness can be a cement wall section or a wooden wall to simulate different detection scenarios. The model trained on the termite dataset corresponding to different wall materials is used to monitor the detection targets of different wall materials. The trained MobileNetV2 models corresponding to different wall materials can be deployed on mobile terminals, and users can be provided with optional detection operations for the target material.

[0033] Furthermore, based on monitoring changes in radar echo energy, the mobile terminal confirmed the specific areas of termite activity identified in the preliminary detection results, including:

[0034] The mean energy value of each image region over a period of time is statistically analyzed, and its variance is calculated as the evaluation value of each region.

[0035] Based on the set energy threshold, the confidence value of areas below the energy threshold is directly set to 0;

[0036] The areas with a confidence value of non-zero represent the active areas of termites.

[0037] Furthermore, the energy threshold is set using an adaptive adjustment strategy:

[0038] Initialize the energy threshold, and based on this initial value, process whether to set the confidence value of each region to 0;

[0039] From the region with a confidence value of 0, the mean value μ of the radar echo energy is statistically analyzed. b and standard deviation σ b ;

[0040] Based on formula T adaptive =μ b +k·σ b Obtain the adjusted energy threshold T adaptive , where k is an adjustment coefficient used to control the strictness of the threshold, a preset value;

[0041] Furthermore, the system continuously monitors the accuracy of the adjusted energy threshold in detecting termite activity during operation to evaluate the detection effect of different values ​​of the adjustment coefficient k, and selects the optimal adjustment coefficient k for the next time period based on the detection effect.

[0042] The technical solution provided by this invention brings at least the following beneficial effects:

[0043] Improving detection accuracy: By combining MIMO radar technology, deep learning algorithms, and radar energy change detection, the accuracy of termite activity detection has been significantly improved. The initial accuracy using the MobileNet model was 87.6%, which increased to 95.0% after incorporating radar energy change detection. This significant improvement greatly enhances the system's reliability and practicality in real-world applications.

[0044] High real-time performance: Utilizing Raspberry Pi and mobile terminals for data processing and model inference, real-time detection and feedback are achieved. The entire process, from scanning data to obtaining termite detection results, has a latency of only 0.2 seconds. This high real-time performance ensures that the system can provide accurate detection results in a short time, which is of great significance for rapid response and appropriate measures.

[0045] High adaptability and user-friendliness: The system design is based on the Android platform, exhibiting excellent mobile device compatibility. Users can easily use the system and record detection locations through a well-designed user interface. This not only facilitates termite detection but also provides convenience for subsequent treatments (such as termite extermination). Users can view the detection results on their mobile phones and take appropriate control measures based on the results. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 The diagram illustrates the structure of an intelligent termite detection system based on MIMO radar technology and deep learning.

[0048] Figure 2 This is a schematic diagram of the main structure of an intelligent termite detection system based on MIMO radar technology and deep learning.

[0049] Figure 3 This is a flowchart illustrating the termite detection process.

[0050] Figure 4 This is a flowchart of the data processing and model training process.

[0051] Figure 5 This is a schematic diagram of a mobile terminal application interface.

[0052] Among them, 1 is the inner box, 2 is the microwave radar equipment, 3 is the Raspberry Pi data processing unit, 4 is the outer box, 5 is the power module, and 6 is the external interface. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be described in detail and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Generally, the components of the embodiments of the present invention described and shown in the accompanying drawings can be arranged and designed using different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present invention.

[0054] The intelligent termite detection system based on MIMO radar technology and deep learning provided in this embodiment of the invention includes a box-type device and a mobile terminal. The box-type device is deployed near the target being detected. Figure 1 As shown, it includes an outer casing 4, an inner casing 1, a microwave radar device 2, a Raspberry Pi data processing unit 3, a power module 5, and external interfaces 6 (such as charging interfaces, USB interfaces, etc.). The external interfaces are located on one side of the outer casing 4 for easy system connection and charging. The microwave radar device 2 is fixed to the bottom of the outer casing 1, with the radar scanning direction downwards. The Raspberry Pi data processing unit 3 and the power module 5 are located in the middle and on one side of the inner casing, respectively, providing data processing and power support for the stable operation of the system. To clearly illustrate the structure, Figure 1 The outer casing 4 is made transparent, meaning that only its outline is displayed without obstructing the view of the internal structure, so that the microwave radar device 2, the Raspberry Pi data processing unit 3, and the power module 5 hidden inside the casing can be shown. Figure 2 This is a schematic diagram of the main structure of the enclosure device. The diagram shows the layout of the internal components. Specifically, the outer enclosure 4 contains the internal microwave radar device 2, Raspberry Pi data processing unit 3, and power module 5. In this embodiment, the microwave radar device 2 is fixed to the bottom of the inner enclosure 1 and is positioned relative to the outer enclosure 4. The microwave radar device 2 is connected to the Raspberry Pi data processing unit 3 via USB connection, and the power module 5 provides stable power to both.

[0055] In this embodiment of the invention, the microwave radar device 2 is used to perform radar scanning on the target to be detected. It adopts a MIMO antenna array structure, which can acquire radar echo data of the scanned area and obtain 2D image data through imaging algorithms. The data is then transmitted to the Raspberry Pi data processing unit 3 via USB connection for data preprocessing (noise filtering and image enhancement) to obtain the data to be detected and transmitted to the mobile terminal via wireless network.

[0056] This mobile terminal is equipped with a termite detection application based on the MobileNetV2 model. It performs target detection on received data to obtain preliminary detection results. Based on monitoring changes in radar echo energy, it confirms the termite activity areas within the preliminary detection results, generates the final termite detection results, and displays them visually on the mobile terminal. Furthermore, the final termite detection results can be pushed to designated users for monitoring and alarm purposes. This embodiment of the invention, through radar energy change monitoring, can more accurately confirm the presence of termite activity because even minute termite movements cause energy fluctuations. By combining energy criteria, the accuracy of termite detection can be effectively improved.

[0057] Typically, a MobileNetV2 model includes: an input layer that receives the input image, i.e., preprocessed 2D image data; a start convolutional layer, a standard 3×3 convolutional layer, for initial feature extraction; a sequence of inverted residual blocks, consisting of multiple inverted residual blocks; a final convolutional layer: a 1x1 point convolution, for generating the final feature map; a global average pooling layer: compressing the feature map into a single vector; and a fully connected layer: for the output of the classification task. Each inverted residual block includes: an expanded 1x1 convolution to increase the number of channels; a depthwise convolution (3x3 or 5x5) with a stride of 1 or 2 for spatial feature extraction; a linear bottleneck 1x1 convolution (squeeze and excite), i.e., a depthwise separable convolution, for dimensionality reduction; skip connections (only when the input and output dimensions are the same); and a fully connected layer (also called the output layer) with a Softmax activation function, used to generate the probability distribution of the binary classification (presence or absence of termites) for each termite detection, thus outputting a confidence value. These confidence scores can be used to assess the model's confidence level for each prediction, allowing users to set judgment thresholds according to their specific needs.

[0058] Figure 3The complete workflow of the termite detection system provided in this embodiment of the invention is described as follows: After system startup, the microwave radar device 2 is initialized by connecting to the Raspberry Pi data processing unit 3, setting radar parameters and establishing a data transmission link. After initialization, the microwave radar device 2 begins scanning the target area, collecting radar signals and 2D image data from within the wall, and transmitting the collected data to the Raspberry Pi for preprocessing in real-time. During this process, the system further monitors radar energy changes. Even slight movements of termites can cause significant changes in the energy value received by the radar. By analyzing these energy changes, the system can more accurately confirm termite activity. The data after energy change detection is input into the MobileNetV2 model for inference. The model classifies the data, determines whether termite activity exists, and outputs the detection results. The detection results are transmitted to the user's mobile phone via a wireless network. The user can view the detection results through a mobile application and obtain detailed visualization information on the location and extent of termite activity. The user can also take further termite control measures based on the detection report provided by the application, such as notifying a professional team for treatment. Furthermore, the application supports recording and storing detection history for convenient user retrieval and review.

[0059] In this embodiment of the invention, see Figure 4 The specific implementation process of training the MobileNetV2 model includes:

[0060] First, radar equipment is used to scan the target area. The collected radar signals and 2D image data undergo a series of preprocessing steps. To ensure data quality, a Butterworth high-pass filter is used to remove low-frequency noise (such as baseline drift below 0.1Hz), and histogram equalization is used to enhance image contrast, making subtle changes in termite activity more clearly visible. The preprocessing stage also applies data augmentation techniques such as contrast enhancement, brightness enhancement, image flipping, and rotation, tailored to the termite's living environment, to improve the model's adaptability to diverse data.

[0061] During the data annotation phase, the collected image data was manually annotated using the open-source tool LabelImg to ensure accurate identification of termite activity areas. The radar imaging area was divided into multiple regions (e.g., 18 regions) for annotation to more accurately locate termites. If sufficient labeled data is available, a pre-trained model can be used for automatic annotation, with manual checks and adjustments to ensure accuracy. To prevent data imbalance, this embodiment divides the dataset into training, validation, and test sets in a 7:1:2 ratio to ensure uniform sample distribution in each subset. The datasets are divided into two types: one collected from cement walls and the other from wooden walls, used to train two different models for users to choose from based on their specific needs.

[0062] In terms of model architecture selection, MobileNetV2 is chosen as the base model. Its depthwise separable convolutions significantly reduce computational cost, making it well-suited for operation on resource-constrained devices. To further enhance the model's feature extraction capabilities, a Coordinate Attention (CA) mechanism is introduced. This mechanism effectively captures information from different spatial locations in the feature map, thereby improving the accuracy of termite activity detection. A channel attention mechanism module, such as an SE module, can be added to the middle of each inverted residual block of the MobileNetV2 model. This SE module includes global average pooling and two fully connected layers. Global average pooling extracts the importance of each channel in the feature map, and then the fully connected layers learn the weights between channels. Finally, the feature maps of different channels are weighted and fused, thus achieving channel attention regulation.

[0063] Regarding hyperparameter settings, the initial learning rate was set to 0.001, and a learning rate scheduler such as ReduceLROnPlateau was used to automatically adjust the learning rate based on the validation set loss to prevent overfitting. The batch size was set to 32 to ensure more stable gradient estimation. Furthermore, during model training, the Adam optimizer was used to accelerate model convergence, and data augmentation techniques such as spatial transformations like rotation, translation, and scaling were applied to further improve the model's generalization ability.

[0064] For model evaluation, a comprehensive assessment was conducted using metrics such as accuracy, precision, recall, and F1 score, with particular attention paid to the number of false positives and false negatives in the confusion matrix to ensure the model's ability to classify boundary samples was optimized. k-fold cross-validation was introduced to verify the model's stability and generalization ability. In the later stages of training, the model weights were further optimized by gradually reducing the learning rate, and an early stopping mechanism was incorporated to prevent overfitting.

[0065] In this embodiment of the invention, the termite dataset is constructed as follows:

[0066] By contacting termite control agencies, we obtained wood samples with diverse morphologies of termite infestation. Using appropriately sized transparent acrylic boxes as the foundation for the experimental environment, a uniformly thick wall material was affixed to the front of the box. The box was then divided into two areas: one for normal wood blocks and the other for termite-infested wood samples, separated by a partition. A suitable amount of subterranean termites was added to the side containing the termite-infested samples. During the experiment, the wall material was changed to either cement sections or wooden walls to simulate different detection scenarios. In addition to collecting data in the constructed environment, real-world data was also collected from termite-infested houses. Furthermore, the front wall was marked with area divisions for later annotation of the model training dataset.

[0067] Once the MobileNetV2 model for termite detection is trained, it is deployed on a mobile device to detect termites on the wall to be inspected.

[0068] After model detection is completed, the system further utilizes an energy threshold algorithm to assist in the judgment of the detection results. Specifically, the radar imaging results are divided into 18 region blocks, and the system returns the energy value of each region through the relevant functions of the radar board. Subsequently, the system calculates the average value of the energy values ​​of each region over a period of time and calculates its variance as the evaluation value of that region. Since experiments have shown that changes in energy values ​​can reflect the presence of biological activity in a region, the system sets an initial threshold T0, which can be set based on the actual application scenario. In this embodiment, it is set to 10. Regions with a confidence value below this threshold are directly set to 0.

[0069] In addition, to eliminate the interference of environmental factors on the judgment, the system dynamically adjusts this energy threshold, and the specific steps include:

[0070] (1) Termite-free area assessment: The system selects areas judged as "termite-free" from the model detection results. These areas are considered to have no biological activity, and their energy values ​​are mainly affected by environmental factors. Next, the system statistically analyzes the energy assessment values ​​of these termite-free areas and calculates their mean value μ. b and standard deviation σ b This is to reflect the normal energy fluctuation range under environmental conditions.

[0071] (2) Adaptive Threshold Adjustment: The system dynamically adjusts the initially set energy threshold T0 based on background energy statistics. The adjusted threshold T... adaptive Calculated using the following formula:

[0072] T adaptive =μ b +k·σ b

[0073] Here, k is an adjustment coefficient used to control the strictness of the threshold. The system selects an appropriate value for k based on the actual application requirements; for example, k can be set to 1 or 2. Subsequently, the system uses the adjusted threshold T. adaptive All regions are re-evaluated. If the energy assessment value of a region is lower than the new threshold, the confidence level of that region is set to 0.

[0074] (3) Verification and Feedback: During operation, the system continuously monitors the impact of the adjusted threshold on the accuracy of termite activity detection. Based on experimental verification of the effects of different k values, the optimal threshold adjustment strategy is selected to balance the false detection rate and the false negative rate. The system will also iteratively optimize the threshold adjustment strategy based on actual detection results to ensure high accuracy and low false alarm rate under various environmental conditions.

[0075] The above processing method analyzes the energy value of termite-free areas and adjusts the detection threshold in real time, enabling the system to dynamically adapt to different environmental backgrounds and thus achieve more accurate and reliable detection of biological activities.

[0076] During the model deployment phase, model quantization was performed for both materials (using TensorFlow Lite) to reduce model size and accelerate inference, ensuring real-time performance and efficiency on mobile devices. Ultimately, through these fine-tuning and optimizations, the termite detection system not only improved detection accuracy but also output the confidence level of each detection result, making the system more reliable and interpretable under different environmental conditions.

[0077] Figure 5 This is a schematic diagram of the termite detection application interface on a mobile terminal according to an embodiment of the present invention, demonstrating the main operating interface when using the intelligent termite detection system. The interface features a title bar at the top, a "Start / Stop" control button on the left, and a "Save" button on the right, allowing users to save the current detection frame during the detection process. A large display area in the center of the interface displays the detection results in real time. The system uses radar image results to render the current area in real time, obtaining the model's detection results within 18 region blocks, and simultaneously outputting the confidence score for each region. The information bar below displays the model's inference time and wall material tabs, while the "History" button allows users to view previously saved data. The overall interface design is simple and clear, centered on user experience, ensuring that users can easily find the functions they need during operation. Figure 5 Users can clearly understand how to use and operate the intelligent termite detection system, thus enabling them to conduct termite detection work more effectively.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

[0079] The above descriptions are merely some embodiments of the present invention. Those skilled in the art can make various modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the scope of protection of the present invention.

Claims

1. An intelligent termite detection system based on MIMO radar technology and deep learning, characterized in that, Includes enclosure equipment and mobile terminals; The enclosure equipment is deployed near the target to detect it using microwave radar and to obtain radar echo signals from inside the target. The enclosure includes an outer enclosure, an inner enclosure, a microwave radar device, a Raspberry Pi data processing unit, a power module, and external interfaces. The microwave radar device is fixed to the bottom of the outer enclosure with its radar scanning direction downward. The Raspberry Pi data processing unit, which is connected to the microwave radar device, is located on the top of the outer enclosure. The power module is used to power the microwave radar device and the Raspberry Pi data processing unit. The outer enclosure of the microwave radar device is also equipped with external interfaces, which serve as power interfaces and data transmission interfaces. The Raspberry Pi data processing unit is used to set the operating parameters of the microwave radar equipment. The microwave radar equipment is used to collect radar echo signals inside the target and send them to the Raspberry Pi data processing unit. The Raspberry Pi data processing unit is used to filter noise from the radar echo signals, then perform imaging processing to obtain the corresponding two-dimensional image data, and perform image enhancement processing to obtain the data to be detected. The data to be detected is then transmitted to the mobile terminal via a wireless network. Based on the preset division of the image region generated by the echo signal, the Raspberry Pi data processing unit calculates the radar energy of each region based on the received echo signal and transmits it to the mobile terminal via a wireless network. The mobile terminal is equipped with a termite detection application based on the MobileNetV2 model, which is used to perform target detection on the received data to be detected and obtain preliminary detection results; and based on the monitoring of changes in radar echo energy, confirm the activity area of ​​termites in the preliminary detection results, generate the final termite detection results, and display them on the mobile terminal.

2. The system as described in claim 1, characterized in that, The termite detection application's user interface on mobile devices includes: Start / Stop Function: Used to start and stop data saving; Real-time monitoring display area: Used to display real-time video or images transmitted by microwave radar equipment, and to mark the confidence level of the corresponding area when termite activity is detected; Data saving function: Used to save the test data and results after the test is completed.

3. The system as described in claim 2, characterized in that, The termite detection application also includes a historical detection record query function in the mobile terminal user interface, allowing users to view and analyze past detection data, which supports filtering and review by time criteria.

4. The system as described in claim 1, characterized in that, Noise filtering uses a high-pass filter to remove background noise signals that do not exceed a specified frequency.

5. The system as described in claim 1, characterized in that, The Raspberry Pi data processing unit uses histogram equalization to perform image enhancement processing.

6. The system as described in claim 1, characterized in that, The MobileNetV2 model consists of an input layer, a starting convolutional layer, an inverted residual block sequence, a convolutional layer before the end, a local average pooling layer, and a fully connected layer based on the Softmax activation function. This layer outputs the termite detection confidence score for each image region of the data to be detected, thus obtaining preliminary detection results.

7. The system as described in claim 1, characterized in that, The termite dataset required for training the MobileNetV2 model is constructed as follows: Obtain wood samples with diverse shapes and forms that have been eaten by termites; In the constructed experimental environment, a wall material of uniform thickness was placed, and two areas were set up behind the wall material. One area was placed with normal wood blocks, and the other area was placed with the obtained wood samples that had been eaten by insects. The two areas were separated by a partition, and a certain number of termites were added to the side where the wood sample was placed. The areas were marked on the wall material of uniform thickness to correspond to the image division of the two-dimensional image obtained by radar echo signal imaging. The enclosure device was deployed in front of a wall material of uniform thickness to collect the termite dataset required for training the MobileNetV2 model. Among them, the MobileNetV2 models trained on termite datasets corresponding to different wall materials are used to monitor the detection targets of different wall materials. The trained MobileNetV2 models corresponding to different wall materials are deployed on mobile terminals, and users are given the option to detect the target material.

8. The system as described in claim 1, characterized in that, Based on monitoring changes in radar echo energy, the mobile terminal confirmed the specific areas of termite activity identified in the preliminary detection results, including: The mean energy value of each image region over a period of time is statistically analyzed and its variance is calculated as the evaluation value of each region. Based on the set energy threshold, the confidence value of areas below the energy threshold is directly set to 0; The areas with a confidence value of non-zero represent the active areas of termites.

9. The system as described in claim 8, characterized in that, The energy threshold is set using an adaptive adjustment strategy: Initialize the energy threshold, and based on this initial value, process whether to set the confidence value of each region to 0; The mean value of radar echo energy is statistically analyzed from the region with a confidence level of 0. and standard deviation ; Based on formula Obtain the adjusted energy threshold ,in, This is an adjustment factor used to control the stringency of the threshold.

10. The system as described in claim 9, characterized in that, During operation, the system continuously monitors the accuracy of the adjusted energy threshold in detecting termite activity to evaluate the adjustment coefficients for different values. Based on the detection results, the adjustment coefficient for the next time period is selected as the optimal result. .

Images