Air-ground collaborative dam termite trace detection method and device and electronic equipment

Abstract

The invention discloses an air-ground collaborative dam termite trace detection method and device and electronic equipment. The method comprises the steps that an unmanned aerial vehicle is used for cruising and scanning the surface of a dam to obtain image information; constructing a three-dimensional map of the dam by using the image information; extracting termite activity trace characteristics on the image information; according to the image information, a detection and recognition model is constructed, and the detection and recognition model is that an efficient detector for a small target is introduced on the basis of yov8; deploying the detection and recognition model on a ground tracked robot, and performing initial detection and recognition on the termite activity trace characteristics by the unmanned aerial vehicle by using the detection and recognition model; the termite traces recognized through initial detection are determined as suspected features, and the position points of the suspected features are determined and calibrated on a map; and the ground tracked robot plans a one-way planning route map to a position point according to the position point calibrated on the three-dimensional map, then navigates to the position point with the suspected feature, and carries out detection confirmation, so that air-ground collaborative integrated dam termite earth surface exposure feature detection is comprehensively realized.

Description

Technical Field

[0001] The present application relates to the technical field of termite control, and in particular to an air-ground coordinated dam termite trace detection method and device, and electronic equipment. Background Art

[0002] Termites build nests, build paths and reproduce on dams, which can damage the internal structure of dam projects and easily induce dangerous situations such as leakage and pitting of water conservancy projects. In serious cases, they may even cause accidents such as dam crossing and embankment collapse, bringing major risks and hidden dangers to the safe operation of reservoir dams, embankments and other projects. Termite detection is the difficulty, focus and pain point of dam termite control technology.

[0003] In order to determine the activities of termites on the dam, especially the location of termites on the dam, the current termite detection method usually adopts manual inspection and calibration. People go to the dam manually to check the damage and activities of soil-dwelling termites on the dam. According to the surface exposed features of termites on the dam (exposed termite features include surface traces of termite activities and termite nest indicators, surface traces include ant roads and mud lines, etc., termite nest indicators include termite nest umbrellas, carbonaria, flying holes, vents, excrement and ant ridges), the position of termite nests is manually determined according to the position of termite exposed features, and the position is marked; when re-inspection is to be carried out, personnel will go to the dam to mark the location to confirm the activities of termites during this period, such as whether the termite damage has been controlled.

[0004] However, the existing termite exposed feature detection methods have the following shortcomings: The surface exposed features of dam termites are attached to the ground, which makes them easily disappear due to external factors such as animal activities, human activities or natural wind. For example, ant paths, mud blankets, flying holes, vents, water lines and fungal indicators are easily destroyed and disappeared by biological activities or rain erosion, resulting in the loss of some exposed termite features during the inspection process. Even during the inspection process, the physical markings of termite exposed features found will disappear due to external factors such as animal activities, human activities or natural wind, resulting in the inability to determine the position of the originally marked termite exposed features during the re-inspection or verification, which has an adverse effect on the calibration of termite indicator positions, the determination of termite nest locations and the understanding of termite activities.

[0005] The detection of exposed features of termites on dams is mainly carried out manually for census and prevention. In summer and autumn every year, experienced personnel are organized to conduct census and prevention operations on the dam site. First, manual inspections are carried out to mark the places where termites appear, and then they are confirmed by experienced masters. After confirmation, appropriate means are used to kill the termites according to their specific circumstances. This method is time-consuming, labor-intensive, inefficient, and has average detection quality.

[0006] The exposure characteristics of dam termites are highly seasonal, mainly concentrated in April to June and September to October each year, and are greatly affected by the climate. The current manual detection cycle is long, the climate is hot during the census, and detection is difficult, making it difficult to conduct round-the-clock inspections. Summary of the invention

[0007] In order to solve the above-mentioned deficiencies, the purpose of the embodiments of the present application is to provide an air-ground collaborative dam termite trace detection method and device, and electronic equipment.

[0008] According to a first aspect of an embodiment of the present application, a method for detecting termite traces in an air-ground coordinated dam is provided, comprising: Use drones to cruise and scan the surface of the dam to obtain image information; constructing a three-dimensional map of the dam using the image information; Extracting termite activity trace features from the image information; According to the image information, a detection and recognition model is constructed. The detection and recognition model is based on yolov8 and introduces an efficient detector SECA-YOLO for small targets. The efficient detector SECA-YOLO includes a feature enhancement module (FEM), a feature fusion module (FFM) and a spatial context perception module (SCAM). The image information is processed by multiple convolution modules on yolov8 to extract key features and aggregate the extracted features; the extracted and aggregated features are processed by the feature enhancement module FEM for multiple feature enhancement of local features; the result of each processing is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after fusion, it is input into the spatial context perception module SCAM to distinguish between the background and the feature image, and finally the required information is extracted; The detection and recognition model is deployed on a ground crawler robot, and the drone uses the detection and recognition model to perform preliminary inspection and recognition of the termite activity trace characteristics; Identifying the termite traces identified in the initial inspection as suspected features, determining the location points of the suspected features and marking them on the three-dimensional map; The ground crawler robot plans a single-line planning route map to the position point according to the position point marked on the three-dimensional map; The tracked robot navigates to the location point of the suspected feature based on the single-line planning route map and performs multi-sensor detection and confirmation.

[0009] According to a second aspect of an embodiment of the present application, there is provided an air-ground coordinated dam termite trace detection device, comprising: An acquisition module is used to obtain image information by using a drone to cruise and scan the surface of the dam; A first construction module is used to construct a three-dimensional map of the dam using the image information; An extraction module, used for extracting the features of termite activity traces from the image information; The second construction module is used to construct a detection and recognition model according to the image information. The detection and recognition model is based on yolov8 and introduces an efficient detector SECA-YOLO for small targets. The efficient detector SECA-YOLO includes a feature enhancement module (FEM), a feature fusion module (FFM) and a spatial context perception module (SCAM). After the image information is processed by multiple convolution modules on yolov8, it enters the feature extraction module to extract key features and aggregates the extracted features; the extracted and aggregated features are processed by the feature enhancement module FEM for multiple feature enhancement of local features; the result of each processing is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after fusion, it is input into the spatial context perception module SCAM to distinguish between the background and the feature image, and finally extract the required information; An initial inspection and recognition module is used to deploy the detection and recognition model on the crawler robot, and the drone uses the detection and recognition model to perform initial inspection and recognition on the characteristics of the termite activity traces; A positioning marking module, used to identify the termite traces identified in the initial inspection as suspected features, determine the location points of the suspected features and mark them on the three-dimensional map; A route planning module, used for the ground crawler robot to plan a single-line planning route map to the position point according to the position point marked on the three-dimensional map; The confirmation module is used for the crawler inspection robot to navigate to the location point of the suspected feature based on the single-line planning route map, and perform multi-sensor detection and confirmation.

[0010] According to a third aspect of an embodiment of the present application, there is provided an electronic device, including: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in the first aspect.

[0011] According to a fourth aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which computer instructions are stored, and when the instructions are executed by a processor, the steps of the method described in the first aspect are implemented.

[0012] The technical solution provided by the embodiments of the present application may have the following beneficial effects: It can be seen from the above embodiments that the surface of the dam is scanned by a drone for cruise to obtain image information; a three-dimensional map of the dam is constructed using the image information; the characteristics of termite activity traces in the image information are extracted; a detection and recognition model is constructed based on the image information, and the detection and recognition model is based on yolov8 and introduces SECA-YOLO, an efficient detector for small targets; the detection and recognition model is deployed on a ground tracked robot, and the drone uses the detection and recognition model to perform an initial inspection and recognition of the characteristics of the termite activity traces; the termite traces identified in the initial inspection are identified as suspected features, and the location points of the suspected features are determined and marked on the three-dimensional map; the ground tracked robot plans a single-line planning route map to the location points based on the marked location points on the three-dimensional map; the tracked robot navigates to the location points of the suspected features based on the single-line planning route map, performs multi-sensor detection and confirmation, and comprehensively realizes the detection of exposed surface features of termites on the dam with air-ground collaborative integration.

[0013] This application adopts a collaborative operation method of drone inspection and ground crawler robot approach confirmation to further improve the efficiency of dam termite inspection, improve detection results, and reduce labor costs, overcoming the problem of insufficient detection accuracy and difficulty in precise positioning caused by single drone operation in similar patents. This application constructs a detection and recognition model based on the image information and deploys the model on a ground crawler robot to form a complete dam termite surface exposed feature detection system, providing an effective and low-cost detection method for dam termite prevention and control.

[0014] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0016] Figure 1 It is a flow chart of a method for detecting termite traces in an open-air and ground-coordinated dam according to an exemplary embodiment.

[0017] Figure 2 It is a flowchart of an improved YOLOV8 algorithm according to an exemplary embodiment.

[0018] Figure 3 It is a schematic structural diagram of a tracked robot according to an exemplary embodiment.

[0019] Figure 4 It is a block diagram of an open-air and ground-coordinated dam termite trace detection device according to an exemplary embodiment. DETAILED DESCRIPTION

[0020] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0021] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0022] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0023] Figure 1 is a flow chart of a method for detecting termite traces in an open space and a dam in coordination with an exemplary embodiment. Figure 1 As shown, the method may include the following steps: S1: Use drones to cruise and scan the surface of the dam to obtain image information; To ensure the safety and integrity of the dam, technicians can use drones to conduct regular patrol scans. This process usually includes the following steps:

[0024] 1. Preliminary preparation: Before you start, you need to set the flight parameters of the drone, such as flight altitude, speed, scanning interval, etc. At the same time, you also need to ensure that the drone's camera and other sensors are in good working condition to ensure image quality.

[0025] 2. Path planning: According to the length, width and terrain characteristics of the dam, one or several optimal flight paths are planned in advance. This step can be achieved through professional flight planning software to ensure that the drone can cover the entire dam surface without leaving any blind spots.

[0026] 3. Takeoff and cruise: After starting the drone, it will automatically fly along the preset flight path. During this process, the high-resolution camera on the drone will continuously take photos of the surface of the dam, or record a video stream in real time. The drone may use oblique photography technology to shoot the dam from multiple angles for later 3D modeling.

[0027] 4. Data collection: The image information captured by the drone will be transmitted to the ground workstation in real time or after the flight.

[0028] In this embodiment, the image information is obtained by a drone performing a cruise scan of the dam to be detected at a low altitude of 3 to 5 meters.

[0029] In this way, drones not only improve the efficiency and accuracy of dam inspections, but also reduce the risks of manual inspections, especially in those hard-to-reach or dangerous areas.

[0030] S2: constructing a three-dimensional map of the dam using the image information; Specifically, the present application uses the image information to construct a three-dimensional map of the dam using SLAM (Simultaneous Localization and Mapping) technology, which is an advanced method that combines real-time positioning and environmental modeling and is very suitable for three-dimensional reconstruction in dynamic environments. The above SLAM technology can not only achieve real-time positioning, but also build an accurate three-dimensional map, which has important application value in safety monitoring, maintenance, and disaster warning of dams.

[0031] S3: extracting termite activity trace features from the image information; Specifically, extracting termite activity trace features from image information is a complex but crucial task, especially in the maintenance and safety monitoring of important infrastructure such as dams. To better understand this process, we can break it down into the following steps:

[0032] 1. Image preprocessing: First, the dam surface images taken by the drone need to be preprocessed, including but not limited to image denoising, brightness and contrast adjustment, etc., to ensure that the image quality meets the needs of subsequent analysis. Preprocessing helps to eliminate interference factors such as uneven lighting and shadows, making the image features more obvious.

[0033] 2. Feature definition: Identify which features can indicate termite activity. These features vary with the seasons and weather, and mainly include: termite holes, mud lines and mud sheets, ant paths, fungi (such as Alternaria japonica and Carbonaceae). Since termite activity often appears on the surface of the dam as the above-mentioned surface features, the image may show texture or color changes that are different from the surrounding environment.

[0034] 3. Image segmentation: Use image segmentation technology to divide the dam surface into several regions, each containing similar texture or color features. This method helps focus on areas where termite activity may exist and reduce the impact of background noise.

[0035] 4. Feature extraction: Apply computer vision techniques, such as edge detection, shape analysis, texture analysis, etc., to extract features of suspected termite activity from images. For example, edge detection algorithms can be used to find termite holes and fungi; texture analysis can help identify mud lines, ant trail size and distribution.

[0036] 5. Machine learning model: Improve the YOLOV8 algorithm model and train the model to identify typical features of termite activity. Collect and annotate the features of termite activity on existing dams to build a dataset, which contains image samples of known termite activity areas as training data.

[0037] 6. Feature classification: The extracted features are input into the trained machine learning model for identification and judgment. Based on the learned knowledge, the model distinguishes which features may be caused by termite activity and which are caused by natural phenomena or other reasons.

[0038] 7. Result verification: For areas of suspected termite activity identified by the model, the crawler robot needs to approach and verify further. The crawler robot uses the high-definition camera it carries to zoom in and view detailed images of these areas. For features that are particularly difficult to identify, the high-definition images are sent back to the ground mobile terminal for manual identification to confirm whether there is termite activity.

[0039] Through the above steps, the trace characteristics of termite activities can be effectively extracted from the images taken by drones, providing technical support for the safe maintenance of the dam.

[0040] S4: constructing a detection and recognition model according to the image information; this step may include the following sub-steps: S41: using labelimg to label the image information and establish a feature training library and a verification library; Specifically, we can follow the steps below: 1. Prepare the dataset: First, a large amount of image data needs to be collected. These images are pictures of the dam surface taken by drones, which contain signs of termite activity or the dam surface under normal conditions.

[0041] The images should cover different time periods, different weather conditions, and different types of dams so that the model can generalize to a variety of scenarios.

[0042] 2. Install and configure LabelImg: Install LabelImg on your computer, a Python-based graphical interface tool for image annotation.

[0043] Install necessary dependent libraries, such as PyQt5 and Pillow.

[0044] 3. Annotate the image: Use LabelImg to open the collected image dataset.

[0045] For each image, carefully observe and determine if there are any signs of termite activity, such as termite holes, mud lines, termite paths, fungus plants, etc.

[0046] Once the suspicious area is found, use LabelImg to draw a rectangular box on the image, mark the area, and assign it a label, such as "diversion port", "mud is mud line", etc.

[0047] For each feature point or area, its position should be marked as accurately as possible, which can improve the accuracy of the training model.

[0048] 4. Data division: The labeled data set is randomly divided into two parts: one part is used as a training set and the other part is used as a validation set. Usually, the ratio can be 80% of the data for training and 20% of the data for validation.

[0049] Make sure that both datasets contain all kinds of labeled categories, so that the model can learn the features of all categories during training.

[0050] 5. Generate TXT file: LabelImg will automatically generate TXT files during the labeling process. These files contain image information as well as the location and category of the labeling.

[0051] TXT files are very important for later model training because they provide the specific target locations that the model needs to learn.

[0052] S42: The established training library and feature library are trained and feature extracted using the improved YOLOV8 algorithm (i.e., SECA-YOLO, an efficient detector for small targets, is introduced based on YOLO V8) to establish a detection and recognition model; Specifically, organize all the labeled TXT files and corresponding image files, and ensure that they are in the correct directory structure to facilitate subsequent training script reading.

[0053] refer to Figure 2 The detection and recognition model given in this embodiment is based on YOLOV8, and introduces an efficient detector SECA-YOLO for small targets. The efficient detector SECA-YOLO includes a feature enhancement module (FEM), a feature fusion module (FFM) and a spatial context perception module (SCAM). The image information is processed by multiple convolution modules on yolov8 to extract key features and aggregate the extracted features; the extracted and aggregated features are processed by the feature enhancement module FEM for three feature enhancements on the local features; the result of each processing is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after the image information is processed by the above two modules, the required image information features are further enhanced. In order to effectively distinguish image features from background and further suppress useless background, the enhanced image information is input into the spatial context perception module SCAM to distinguish the background from the feature image, and finally extract the required information; after the above-mentioned yolov8 improved algorithm introducing SECA-YOLO, the final image information is further feature processed and output by yolo-head to obtain the target information to be detected and marked.

[0054] The mathematical expression of the feature enhancement module FEM is: (1) in, , , and They represent standard convolution operations with kernel sizes of 1×1, 1×3, 3×1, and 3×3 respectively; represents a dilated convolution operation with a dilation rate of 5; Cat(·) is the feature map concatenation operation; ⊕ represents the element addition operation of the feature map; F is the input feature map; W1, W2, and W3 represent the output feature maps of the first three branches after standard convolution and dilated convolution; Y is the output characteristic map of the FEM method.

[0055] The mathematical expression of the feature fusion module (FFM) is: (2) in, represents the trainable weight in the i-th feature map; represents the trainable weight of the jth channel; is the number of channels in the i-th feature map; m represents the total number of channels after connection; ε is the adjustment coefficient.

[0056] The mathematical expression of the spatial context awareness module (SCAM) is: (3) In the formula, and Respectively represent the input and output of the jth pixel in the i-th level feature map; Indicates the total number of pixels; and is the linear transformation matrix used to project the feature map, which is simplified by 1×1 convolution; avg(·) and max(·) implement GAP and GMP respectively.

[0057] During the training process, it is necessary to monitor the changing trend of the loss function to ensure the convergence of the model.

[0058] Verify and optimize the model: Test the trained model on the validation set and evaluate its performance metrics such as accuracy, recall, F1 score, etc.

[0059] If the model performs poorly, you need to adjust hyperparameters or add more labeled data to improve the model.

[0060] Based on the results on the validation set, the model is continuously optimized until a satisfactory performance level is achieved.

[0061] Through the above steps, we can use LabelImg to effectively create a training library and a validation library containing termite activity features, and then train a model that can automatically detect signs of termite activity. This will greatly improve the efficiency of dam management, reduce the workload of manual inspections, and detect potential risks earlier.

[0062] S5: deploying the detection and recognition model on the ground crawler robot, and the drone using the detection and recognition model to perform preliminary inspection and recognition of the termite activity trace features; Specifically, the drone conducts a low-altitude (2 to 3 meters) cruise scan of the dam to be inspected to obtain image data information on the dam surface, and imports the acquired information into the detection and recognition model for initial inspection and identification of exposed surface features of termites.

[0063] The crawler inspection robot can adopt the existing one, and on this basis, install the required sensors. Figure 3As shown in the figure, a tracked inspection robot is shown, wherein 1 is a drone in the cabin at the top of the robot, which can realize functions such as drone shutdown, charging, and fault diagnosis, and the drone can realize the initial inspection of dam composition and termite traces; 2 is a millimeter wave radar, which is mainly used for the autonomous navigation process of the tracked robot to realize intelligent obstacle avoidance, adjust the travel route and repair defects in the unmanned structure map; 3 is a drive motor, which provides power for the tracked robot; 4 is a track, which realizes the obstacle crossing and walking of the tracked robot; 5, 6, and 7 are temperature sensors, carbon dioxide sensors, and visual sensors, respectively, which are arranged in a staggered manner according to specific conditions, and are used for multi-sensor fusion detection of termite traces to realize accurate confirmation of termite traces; 8 is a foldable tracked robot support arm, which can dynamically adjust the height and the center of gravity of the tracked robot according to the slope of the dam, the vegetation coverage on the dam surface, and the obstacles on the dam surface, so as to enhance the stability of the tracked robot and the maneuverability on large slopes, and further improve its off-road performance.

[0064] S6: Identify the termite traces identified in the initial inspection as suspected features, determine the location points of the suspected features and mark them on the three-dimensional map.

[0065] Specifically, the coordinate information of the exposed surface features suspected of termites obtained after the initial inspection by the drone is extracted and marked on the constructed three-dimensional map of the dam.

[0066] S7: the ground crawler robot plans a navigation route map to the location point according to the location point marked on the three-dimensional map; Specifically, the path planning algorithm is used to optimize the location of the suspected points marked on the three-dimensional map, and the planned navigation route map is sent to the tracked robot, which then begins to autonomously navigate to the location points on the route map.

[0067] S8: The tracked inspection robot navigates to the location of the suspected feature based on the single-line planning route map and performs multi-sensor detection and confirmation.

[0068] Specifically, the crawler robot carries high-definition cameras, temperature sensors, carbon dioxide sensors and other types of sensors to collect relevant data after it reaches the suspected feature location, and compares the collected data information with the data information in the model to confirm whether it is a real termite surface feature. After confirmation, the location and image information are uploaded to the cloud platform. The cloud platform determines the prevention and control plan based on the actual situation and sends it to various local prevention and control departments.

[0069] It can be seen from the above embodiments that the present application builds an air-ground collaborative platform, adopts the strategy of initial inspection by drones and re-inspection by tracked robots with multi-sensor fusion, improves the deep learning algorithm to identify the exposed features of termites on the surface, and focuses on solving the problems of time-consuming, labor-intensive, high-cost, and greatly affected by climate and environment in the process of embankment termite surveys.

[0070] Corresponding to the aforementioned embodiment of the method for detecting termite traces in an open space and a ground in coordination with each other, the present application also provides an embodiment of a device for detecting termite traces in an open space and a ground in coordination with each other.

[0071] Figure 4 1 is a block diagram of a device for detecting termite traces in an open space and a dam in coordination with an exemplary embodiment. Figure 4 , the device comprises:

[0072] Acquisition module 1, used to obtain image information by using a drone to cruise and scan the surface of the dam; A first construction module 2 is used to construct a three-dimensional map of the dam using the image information; Extraction module 3, used to extract the trace features of termite activity from the image information; The second construction module 4 is used to construct a detection and recognition model according to the image information. The detection and recognition model is based on yolov8 and introduces an efficient detector SECA-YOLO for small targets. The efficient detector SECA-YOLO includes a feature enhancement module (FEM), a feature fusion module (FFM) and a spatial context perception module (SCAM). The image information is processed by multiple convolution modules on yolov8 to extract key features and aggregate the extracted features; the extracted and aggregated features are processed by the feature enhancement module FEM for multiple feature enhancement of local features; the result of each processing is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after fusion, it is input into the spatial context perception module SCAM to distinguish between the background and the feature image, and finally extract the required information; The initial inspection and recognition module 5 is used to deploy the detection and recognition model on the crawler robot, and the drone uses the detection and recognition model to perform initial inspection and recognition on the characteristics of the termite activity traces; A positioning marking module 6 is used to identify the termite traces identified in the initial inspection as suspected features, determine the location points of the suspected features and mark them on the three-dimensional map; Route planning module 7, used for the ground crawler robot to plan a single-line planning route map to the position point according to the position point marked on the three-dimensional map; The confirmation module 8 is used for the tracked inspection robot to navigate to the location point of the suspected feature based on the single-line planning route map, and perform multi-sensor detection and confirmation.

[0073] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0074] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can refer to the partial description of the method embodiment. The device embodiment described above is only schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present application scheme. A person of ordinary skill in the art can understand and implement it without paying any creative work.

[0075] Correspondingly, the present application also provides an electronic device, including: one or more processors; a memory for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned air-ground collaborative dam termite trace detection method.

[0076] Correspondingly, the present application also provides a computer-readable storage medium on which computer instructions are stored, and when the instructions are executed by a processor, the above-mentioned air-ground collaborative dam termite trace detection method is implemented.

[0077] Those skilled in the art will readily appreciate other embodiments of the present application after considering the description and practicing the contents disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The description and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the claims.

[0078] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A method for detecting termite traces in open space and dam, characterized in that: include: Use drones to cruise and scan the surface of the dam to obtain image information; constructing a three-dimensional map of the dam using the image information; Extracting termite activity trace features from the image information; According to the image information, a detection and recognition model is constructed. The detection and recognition model is based on yolov8 and introduces SECA-YOLO, an efficient detector for small targets. The efficient detector SECA-YOLO includes a feature enhancement module, a feature fusion module and a spatial context perception module. The image information is processed by multiple convolution modules on yolov8 to extract key features and aggregate the extracted features. The extracted and aggregated features are processed by the feature enhancement module FEM to perform multiple feature enhancement processes on the local features; the result of each process is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after fusion, it is input into the spatial context perception module SCAM to distinguish the background and feature images, and finally extract the required information; The detection and recognition model is deployed on a ground crawler robot, and the drone uses the detection and recognition model to perform preliminary inspection and recognition of the termite activity trace features; Identifying the termite traces identified in the initial inspection as suspected features, determining the location points of the suspected features and marking them on the three-dimensional map; The ground crawler robot plans a single-line planning route map to the position point according to the position point marked on the three-dimensional map; The tracked robot navigates to the location point of the suspected feature based on the single-line planning route map and performs multi-sensor detection and confirmation.

2. The method according to claim 1, characterized in that The image information is obtained by a drone performing a cruise scan of the dam to be detected at a low altitude of 3 to 5 meters.

3. The method according to claim 1, characterized in that Using the image information to construct a three-dimensional map of the dam includes: The image information is used to construct a three-dimensional map of the dam using Slam technology.

4. The method according to claim 1, characterized in that: According to the image information, a detection and recognition model is constructed, including: S41: using labelimg to label the image information and establish a feature training library and a verification library; S42: Using the improved yolov8 algorithm to train and extract features from the established training library and feature library, a detection and recognition model is established.

5. The method according to claim 1, characterized in that: The mathematical expression of the feature enhancement module is: （1） in, , , and They represent standard convolution operations with kernel sizes of 1×1, 1×3, 3×1, and 3×3 respectively; represents the dilated convolution operation; Cat(·) is the feature map concatenation operation; ⊕ represents the element addition operation of the feature map; F is the input feature map; W1, W2, and W3 represent the output feature maps of the first three branches after standard convolution and dilated convolution; Y is the output characteristic map of the FEM method.

6. The method according to claim 1, characterized in that The mathematical expression of the feature fusion module is: （2） in, represents the trainable weight in the i-th feature map; represents the trainable weight of the jth channel; is the number of channels in the i-th feature map; m represents the total number of channels after connection; ε is the adjustment coefficient.

7. The method according to claim 1, characterized in that The mathematical expression of the spatial context perception module is: （3） In the formula, and Respectively represent the input and output of the jth pixel in the i-th level feature map; Indicates the total number of pixels; and is the linear transformation matrix used to project the feature map, which is simplified by 1×1 convolution; avg(·) and max(·) implement GAP and GMP respectively.

8. A device for detecting termite traces in open space and dam, characterized in that: include: An acquisition module is used to obtain image information by using a drone to cruise and scan the surface of the dam; A first construction module is used to construct a three-dimensional map of the dam using the image information; An extraction module, used for extracting the features of termite activity traces from the image information; The second construction module is used to construct a detection and recognition model according to the image information. The detection and recognition model is based on yolov8 and introduces SECA-YOLO, an efficient detector for small targets. The efficient detector SECA-YOLO includes a feature enhancement module (FEM), a feature fusion module (FFM) and a spatial context awareness module (SCAM). After the image information is processed by multiple convolution modules on yolov8, it enters the feature extraction module to extract key features, and the extracted features are aggregated; The extracted and aggregated features are processed by the feature enhancement module FEM to perform multiple feature enhancement processes on the local features; the result of each process is input into the feature fusion module FFM for secondary weighted feature extraction and fusion; after fusion, it is input into the spatial context perception module SCAM to distinguish the background and feature images, and finally extract the required information; An initial inspection and recognition module is used to deploy the detection and recognition model on the crawler robot, and the drone uses the detection and recognition model to perform initial inspection and recognition on the characteristics of the termite activity traces; A positioning and marking module, used to identify the termite traces identified in the initial inspection as suspected features, determine the location points of the suspected features and mark them on the three-dimensional map; A route planning module, used for the ground crawler robot to plan a single-line planning route map to the position point according to the position point marked on the three-dimensional map; The confirmation module is used for the crawler inspection robot to navigate to the location point of the suspected feature based on the single-line planning route map, and perform multi-sensor detection and confirmation.

9. An electronic device, characterized in that: include: one or more processors; A memory for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instruction is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images