Antenna detection method, device and equipment based on improved YOLO v10 and medium

Abstract

The invention provides an antenna detection method, device and equipment based on improved YOLO v10, and a medium, and belongs to the technical field of target detection, and the method comprises the steps: obtaining a preprocessed satellite antenna image data set; constructing an improved YOLO v10 network structure, and performing learning training on the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model; and inputting a to-be-identified satellite antenna image into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the to-be-identified satellite antenna image. Through the antenna identification mode completed based on the single-stage detection YOLO algorithm, the detection speed of the satellite antenna can be improved; and the Poisson fusion principle and the traditional image augmentation method are utilized to enhance the data, so that the number of data sets can be increased, and the accuracy and generalization ability of the target detection model can be improved.

Description

Technical Field

[0001] The present application relates to the field of target detection technology, and more specifically, to an antenna detection method, apparatus, device and medium based on improved YOLO v10. Background Art

[0002] The illegal use of satellite antennas not only affects the beauty of the urban environment, but also poses safety hazards and unstable factors such as affecting the city appearance, non-standard installation technology, interfering with viewing signals, prone to lightning accidents, spreading illegal content, and affecting physical health. Therefore, it is necessary to remove illegal satellite antennas. One of the most important links is to obtain the distribution location of satellite antennas so that relevant personnel can accurately remove them. The existing means of screening illegal satellite antennas can be roughly divided into the following three ways: (1) Relying on local volunteers and grid members to post satellite antenna management posters in key communities, and use electronic display screens, banners, and mobile loudspeakers to promote and guide users who install satellite antennas to consciously abide by the law and report in a timely manner. (2) Through inspection personnel, a survey and registration of the jurisdiction area is carried out to find out the number and location of satellite antennas. Through door-to-door visits, a carpet-style investigation is carried out in residential communities, street shops, trading markets, etc. to check whether there is any illegal satellite antenna installed. (3) Use "drone" inspection methods, introduce drones to shoot images for a survey, and use the personal judgment of drone pilots to determine whether there are illegal satellite antennas during the drone flight.

[0003] As the number of buildings grows exponentially, manual on-site inspections will consume a lot of manpower and material resources. Secondly, satellite antennas are usually installed at high places, which will pose a threat to the lives of inspectors. When using drones to take close-up photos of buildings, the shooting height cannot be too high, otherwise the image quality will be low, making it difficult to identify later. As the shooting height of the drone decreases, the shooting area will become smaller, resulting in an increase in the number of photos taken. Relying on the senses and experience of professionals for identification, this manual identification method is cumbersome, inefficient, and subjective, and often results in missed judgments and misjudgments. Summary of the invention

[0004] In view of this, the purpose of this application is to provide an antenna detection method, device, equipment and medium based on improved YOLO v10, which can complete the antenna identification of drone aerial photography based on the single-stage detection YOLO algorithm, which is more accurate and efficient.

[0005] In a first aspect, the present application provides an antenna detection method based on improved YOLO v10, the method comprising the following steps:

[0006] Acquire a preprocessed satellite antenna image dataset; wherein the preprocessed satellite antenna image dataset includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images;

[0007] An improved YOLO v10 network structure is constructed, and the improved YOLO v10 network structure is trained based on the satellite antenna image data set to obtain a satellite antenna target detection model; wherein the improved YOLO v10 network structure replaces the PSA module in the original YOLO v10 network structure with a Double Attention module;

[0008] The satellite antenna image to be identified is obtained, and the satellite antenna image to be identified is input into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

[0009] In a possible implementation manner, the step of obtaining the preprocessed satellite antenna image dataset includes the following steps:

[0010] Acquire a plurality of satellite antenna images, and pre-process the acquired plurality of satellite antenna images to filter out images that do not contain satellite antennas;

[0011] Performing image augmentation processing on the plurality of pre-processed satellite antenna images to obtain a satellite antenna image set; wherein the image augmentation processing includes one or more of salt and pepper noise, Gaussian noise, Gaussian blur, brightness enhancement, and Poisson fusion;

[0012] Data annotation is performed on each satellite antenna image in the obtained satellite antenna image set, and a satellite antenna image data set is obtained; wherein the data annotation includes the category of the satellite antenna and a bounding box in the image.

[0013] In a possible implementation, a plurality of satellite antenna images are acquired by taking aerial photos of the building using a drone.

[0014] In a possible implementation, Poisson fusion is performed in the following manner, including the following steps:

[0015] Import the target satellite antenna image and background image;

[0016] Replacing pixels at corresponding positions of the background image with pixels of the target satellite antenna image, and calculating the divergence of the target satellite antenna image;

[0017] Iteratively solving the Poisson equation using the divergence and the values ​​of the pixels corresponding to the outermost part of the target satellite antenna image to obtain the pixel values ​​of the fused target satellite antenna image;

[0018] The pixel values ​​of the target position in the background image are replaced by the pixel values ​​of the fused target satellite antenna image to obtain a final fused image.

[0019] In a possible implementation, the improved YOLO v10 network structure includes a backbone network, a neck and a prediction head, the backbone network adopts CSPNet; the neck adopts a PAN structure; the prediction head includes a one-to-many prediction head and a one-to-one prediction head.

[0020] In a possible implementation, the learning and training of the improved YOLOv10 network structure based on the satellite antenna image dataset to obtain a satellite antenna target detection model includes the following steps:

[0021] Dividing the satellite antenna image data set into a satellite antenna image data training set and a satellite antenna image data test set according to a preset ratio;

[0022] Training the constructed improved YOLO v10 network structure based on the satellite antenna image data training set to obtain an initial satellite antenna target detection model;

[0023] Inputting the satellite antenna images in the satellite antenna image data test set into the initial satellite antenna target detection model to obtain a detection result, and comparing the detection result with the annotated actual result, and verifying the detection accuracy of the initial satellite antenna target detection model according to the comparison result;

[0024] If the detection accuracy of the initial satellite antenna target detection model meets the preset requirements, the initial satellite antenna target detection model is used as the trained satellite antenna target detection model.

[0025] In a possible implementation, the improved YOLO v10 network structure constructed is trained on a tensorfow system based on the satellite antenna image data training set to obtain an initial satellite antenna target detection model; the system parameters set by the tensorfow system include an initial learning rate and an iteration parameter of the initial learning rate.

[0026] In a second aspect, the present application provides an antenna detection device based on improved YOLO v10, the device comprising:

[0027] An acquisition module, used to acquire a preprocessed satellite antenna image data set; wherein the preprocessed satellite antenna image data set includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images;

[0028] A training module is used to construct an improved YOLO v10 network structure, and to perform learning and training on the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model; wherein the improved YOLO v10 network structure replaces the PSA module in the original YOLO v10 network structure with a DoubleAttention module;

[0029] The detection module is used to obtain the satellite antenna image to be identified, and input the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

[0030] In a third aspect, the present application provides an electronic device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory through the bus, and when the machine-readable instructions are executed by the processor, the steps of the antenna detection method based on the improved YOLO v10 as described in the first aspect are performed.

[0031] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the antenna detection method based on the improved YOLO v10 as described in the first aspect are executed.

[0032] The present embodiment provides an antenna detection method, device, equipment and medium based on improved YOLO v10, which obtains a preprocessed satellite antenna image dataset; constructs an improved YOLO v10 network structure, and learns and trains the improved YOLO v10 network structure based on the satellite antenna image dataset to obtain a satellite antenna target detection model; obtains a satellite antenna image to be identified, and inputs the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified. Through this antenna recognition method based on the single-stage detection YOLO algorithm, the detection speed of the satellite antenna can be improved; and in other embodiments, the data is enhanced using the Poisson fusion principle and the traditional image augmentation method, which can not only increase the number of data sets, but also improve the accuracy and generalization ability of the target detection model. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 It is a flowchart of an antenna detection method based on improved YOLO v10 described in an embodiment of the present application;

[0035] Figure 2 A schematic diagram of multiple satellite antenna images in one embodiment of the present application;

[0036] Figure 3 This is a schematic diagram of Poisson fusion principle in one embodiment of the present application;

[0037] Figure 4 A schematic diagram of multiple satellite antenna images after image augmentation processing in one embodiment of the present application;

[0038] Figure 5 A schematic diagram of an improved YOLO v10 network structure in an embodiment of the present application;

[0039] Figure 6 This is a schematic diagram of the recognition effect output by the satellite antenna target detection model described in one embodiment of the present application;

[0040] Figure 7 It is a structural block diagram of an antenna detection device based on improved YOLO v10 described in an embodiment of the present application;

[0041] Figure 8 It is a structural block diagram of an electronic device described in one embodiment of the present application. DETAILED DESCRIPTION

[0042] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of explanation and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn in real proportion. The flowchart used in this application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can be implemented out of sequence, and the steps without logical context can be reversed in order or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart under the guidance of the content of the present application, or remove one or more operations from the flowchart.

[0043] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application usually described and marked in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0044] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.

[0045] Based on the technical problems raised by the background technology, the present application provides an antenna detection method, device, equipment and medium based on the improved YOLO v10, which can complete the antenna identification of drone aerial photography based on the single-stage detection YOLO algorithm, which is more accurate and efficient.

[0046] See the instruction manual Figure 1 In one embodiment, the present application provides an antenna detection method based on improved YOLO v10, comprising the following steps:

[0047] S1. Obtain a preprocessed satellite antenna image dataset; wherein the preprocessed satellite antenna image dataset includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images;

[0048] S2. constructing an improved YOLO v10 network structure, and learning and training the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model; wherein the improved YOLOv10 network structure replaces the PSA module in the original YOLO v10 network structure with a Double Attention module;

[0049] S3. Acquire a satellite antenna image to be identified, and input the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

[0050] Specifically, in step S1, since there is no antenna dataset in the currently public dataset, it is necessary to create a satellite antenna target detection dataset, that is, a satellite antenna image dataset. Figure 2, a plurality of satellite antenna images are obtained by high-altitude photography with a drone, and the obtained plurality of satellite antenna images are preprocessed to remove the images that do not contain satellite antennas. Since the satellite antenna images are taken by drones in real scenes, the amount of data is small, and the effect of the target detection model is related to the amount of data. Therefore, it is necessary to perform data enhancement on the satellite antenna images, which can also improve the generalization ability of the model.

[0051] The methods used in this application for data enhancement of satellite antenna images are divided into two types: 1) image enhancement technology based on traditional methods; 2) image enhancement technology based on Poisson fusion. Traditional image enhancement technologies include salt and pepper noise, Gaussian noise, Gaussian blur and brightness enhancement methods; the salt and pepper noise can highlight the edges and details of the image to a certain extent; the Gaussian noise can simulate various random interferences in the real world to a certain extent, making the image closer to the real scene; the Gaussian blur can be used to blur the background of the image and highlight the subject; the brightness enhancement can improve the contrast of the image and highlight the importance of the area; the specific execution steps should be technical means well known to those skilled in the art, and will not be repeated here. The following only focuses on the image enhancement technology based on Poisson fusion. The basic idea of ​​the fusion algorithm is to use the concept of the field to perform modular interpolation processing and reconstruct the pixel values ​​of the fused part in the area, so as to achieve seamless fusion between the two images.

[0052] See the instruction manual Figure 3 , g is the scene in the image, v is the gradient field of g, S is the fused image domain, and Ω is the scene covered by S. is its boundary, f * is a known scalar function, which represents the pixel value outside Ω, and f is an unknown scalar function, which represents the pixel value inside Ω. In order to achieve a seamless transition without gaps, the gradient value inside Ω should be as small as possible, thus converting the image seamless fusion process into a gradient field minimization problem. Among them, f can be solved using formula (2-1):

[0053]

[0054] in, is the gradient operator, function In the equation, the minimum value of F satisfies the Euler-Lagrange equation, so its solution can be expressed by the Laplace equation, that is, formula (2-2):

[0055]

[0056] in, is the Laplace operator.

[0057] In actual application simulation, the above method may have the defect of over-smoothing, resulting in gradient inaccuracy, which leads to unsatisfactory fusion effect. The Poisson equation uses the gradient field v of g as the guide field to cleverly solve this problem. The purpose of finding the guide field is to make the gradient of f closer to the gradient of g, while maintaining the image detail information to the maximum extent, ensuring that there is no obvious excessive trace at the boundary. After optimization, the formula (2-3) is used:

[0058]

[0059] Substituting into the Euler-Lagrange equation again, the result is shown in formula (2-4):

[0060]

[0061] Among them, div(v) represents the divergence of the gradient field v. This formula is the mathematical basis of Poisson fusion technology. In the fusion process of the target image and the background image, the gradient field of the target image is used as a guidance field to achieve the fusion of the two images. Therefore, the image fusion problem can be abstracted as the solution of the above discretized Poisson equation. From the experimental principle, it is not difficult to see that Poisson fusion actually uses the boundary of the background image to achieve seamless fusion, and combines the gradient of the original image, that is, the target image, that is, the texture change to obtain the merged image. Because the image is a discrete value, the divergence of the image can be obtained by convolving the Laplace operator with the image. F represents the divergence of the image, as shown in formula (2-5):

[0062] F(i,j)=Δf=f(i-1,j)+f(i+1,j)+f(i,j-1)+f(i,j+1)-4*f(i,j) (2-5)

[0063] According to the divergence and Poisson's equation, the fused pixel value can be solved as shown in formula (2-6):

[0064] f1(i,j)=(f(i-1,j)+f(i+1,j)+f(i,j-1)+f(i,j+1)-f(i,j) / 4) (2-6)

[0065] Furthermore, when Poisson fusion is performed in the present application, the following steps are included: importing a target satellite antenna image and a background image; replacing pixels at corresponding positions of the background image with pixels of the target satellite antenna image, and calculating the divergence of the target satellite antenna image; iteratively solving the Poisson equation using the divergence and the values ​​of pixels corresponding to the outermost portion of the target satellite antenna image to obtain pixel values ​​of the fused target satellite antenna image; replacing the pixel values ​​of the target position in the background image with the obtained pixel values ​​of the fused target satellite antenna image to obtain a final fused image.

[0066] The schematic diagram of image augmentation processing of the plurality of satellite antenna images based on salt and pepper noise, Gaussian noise, Gaussian blur, brightness enhancement, and Poisson fusion can be found in the attached manual. Figure 4 .

[0067] In addition, after the satellite antenna image is data enhanced, it is also necessary to perform data annotation. The data annotation is the core link of target detection data collection and annotation, and its purpose is to mark and classify target objects in images or videos. The accuracy and completeness of data annotation directly affect the effect of subsequent target detection model training. The data annotation of this application is divided into the following steps: use visual standard tools for annotation, such as Labelling, BoundingBox, etc.; define the annotation specifications. Since this application is only used to detect illegal satellite antennas, the target type is antenna and the annotation format is a square box; professionals annotate the antennas in the image or video one by one, mark the position and category of the antenna boundary box, and design the model basic data set in a one-to-one format of png-xml; review the annotation results to ensure the accuracy and completeness of the annotation data, and correct or re-annotate the annotation data that does not meet the requirements.

[0068] Finally, the satellite antenna images after data enhancement processing and the data annotations of each satellite antenna image are unified as the satellite antenna image dataset for subsequent target detection model training.

[0069] In step S2, the present application adopts an improved YOLO v10 network structure, which mainly includes a backbone network, a neck and a prediction head. The backbone network uses an enhanced version of CSPNet to extract image features. CSPNet (Cross Stage Partial Network) improves gradient propagation and reduces the amount of calculation by separating the gradient flow of feature maps; the neck adopts a PAN (Path Aggregation Network) structure to aggregate features of different scales. The PAN structure effectively realizes multi-scale feature fusion through bottom-up path enhancement, and improves the model's detection ability for targets of different scales; the prediction head includes a one-to-many prediction head and a one-to-one prediction head, wherein the one-to-many prediction head generates multiple predictions for each object during the training process to provide rich supervision signals, thereby improving the accuracy of learning; the one-to-one prediction head generates an optimal prediction for each object in the reasoning stage, without the need for non-maximum suppression (NMS) operations, thereby reducing the amount of calculation and delay, and improving the reasoning efficiency. In addition, large kernel convolution and partial self-attention modules are introduced to improve performance without significantly increasing the computational cost. A consistent double allocation strategy is used to eliminate the need for NMS and further reduce inference latency. In addition, the PSA module (PyramidSplit Attention) is replaced with the Double Attention module based on the YOLO v10 network model to obtain an improved YOLO v10 network structure. The illegal satellite antenna detection network structure based on the improved YOLO v10 is shown in the attached manual. Figure 5 shown.

[0070] In the field of object detection, adjusting model parameters is a key step to improve model performance and accuracy. The main parameters include learning rate, training batch, optimizer, Confidence threshold, Intersection over Union threshold, and number of iterations. Among them, the learning rate determines the step size of the model weight update and is a hyperparameter used by the optimizer to update the model parameters. The learning rate scheduler (such as warm-up, step decay, cosine decay, etc.) can be used to dynamically adjust the learning rate during the training process; the size of the data training batch determines the amount of data used for training in each iteration. A larger batch size can provide more information to update the model weights, but it may also cause insufficient memory. It is necessary to select a suitable batch size based on hardware resources (such as GPU memory); the optimizer is used to update the model weights to minimize the loss function. Commonly used optimizers include SGD (stochastic gradient descent), Adam, etc.; the Conf threshold is an indicator used to determine whether the detection box contains an object. Each detection box is accompanied by a confidence, which indicates the probability of containing an object in the detection box. When the confidence is greater than the Conf threshold, it is considered that there is an object in the detection box; the IoU threshold is an indicator used to determine the degree of overlap between the detection box and the real object box. IoU is calculated by dividing the intersection area of ​​the detection box and the real object box by their union area. When the calculated IoU is greater than the IoU threshold, the detection box is artificially matched with the real object box. Iteration is used to calculate the loss function of the model and update the weight value to minimize the value of the loss function until the model converges. At the same time, to prevent overfitting, the number of iterations should not be set too large.

[0071] In one embodiment, the improved YOLO v10 network structure is learned and trained based on the satellite antenna image data set to obtain a satellite antenna target detection network, including the following steps: dividing the satellite antenna image data set into a satellite antenna image data training set and a satellite antenna image data test set according to a preset ratio; training the constructed improved YOLO v10 network structure based on the satellite antenna image data training set to obtain an initial satellite antenna target detection network; inputting the satellite antenna images in the satellite antenna image data test set into the initial satellite antenna target detection network to obtain a detection result, and comparing the detection result with the marked actual result, and verifying the detection accuracy of the initial satellite antenna target detection network according to the comparison result; if the detection accuracy of the initial satellite antenna target detection network meets the preset requirements, the initial satellite antenna target detection network is used as the trained satellite antenna target detection network. For example, the satellite antenna image dataset is divided into a training set and a test set in a ratio of 8:2, and a tensorfow system (a symbolic mathematical system based on data flow programming) is used as the backend of the improved YOLO v10 network structure for learning and training, wherein the initial learning rate of the tensorfow system is set to 0.0001, the batch size is 32, and the number of iterations is 200.

[0072] In step S3, i.e. in actual application, the trained satellite antenna target detection model is deployed to a real scene, so that it can recognize and detect new, unseen antenna images, and mark the position of the satellite antenna in the satellite antenna image to be identified. The recognition results can be found in the appendix of the manual. Figure 6 .

[0073] The present application provides an antenna detection method based on improved YOLO v10. Compared with other algorithms, the single-stage detection method YOLO algorithm has a small model framework, flexible deployment, and fast detection rate. Therefore, it is applied to antenna detection in aerial images to improve detection efficiency. In the present application, the satellite antenna images collected in real scenes are expanded by utilizing the Poisson fusion principle and traditional image augmentation methods, which are used for training the improved YOLO v10 network structure, thereby improving the accuracy and generalization ability of the satellite antenna target detection model.

[0074] As the instruction manual Figure 7 As shown, the present application also provides an antenna detection device based on improved YOLO v10, the device comprising:

[0075] The acquisition module 701 is used to acquire a pre-processed satellite antenna image dataset; wherein the pre-processed satellite antenna image dataset includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images;

[0076] A training module 702 is used to construct an improved YOLO v10 network structure, and to perform learning and training on the improved YOLO v10 network structure based on the satellite antenna image dataset to obtain a satellite antenna target detection model; wherein the improved YOLO v10 network structure replaces the PSA module in the original YOLO v10 network structure with a DoubleAttention module;

[0077] The detection module 703 is used to obtain the satellite antenna image to be identified, and input the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

[0078] In some embodiments, the acquisition module 701 uses a drone to take aerial photos of the building to obtain a preprocessed satellite antenna image data set, including: obtaining multiple satellite antenna images, and preprocessing the multiple satellite antenna images to filter out images that do not contain satellite antennas; performing image augmentation processing on the preprocessed multiple satellite antenna images to obtain a satellite antenna image set; wherein the image augmentation processing includes one or more of salt and pepper noise, Gaussian noise, Gaussian blur, brightness enhancement, and Poisson fusion; performing data annotation on each satellite antenna image in the obtained satellite antenna image set to obtain a satellite antenna image data set; wherein the data annotation includes the category of the satellite antenna and the bounding box in the image.

[0079] In some embodiments, the acquisition module 701 performs Poisson fusion in the following manner, including: importing a target satellite antenna image and a background image; replacing pixels at corresponding positions of the background image with pixels of the target satellite antenna image, and calculating the divergence of the target satellite antenna image; iteratively solving the Poisson equation using the divergence and the values ​​of pixels corresponding to the outermost portion of the target satellite antenna image to obtain pixel values ​​of the fused target satellite antenna image; replacing the pixel values ​​of the target position in the background image with the obtained pixel values ​​of the fused target satellite antenna image to obtain a final fused image.

[0080] In some embodiments, the improved YOLO v10 network structure includes a backbone network, a neck and a prediction head, the backbone network adopts CSPNet; the neck adopts a PAN structure; the prediction head includes a one-to-many prediction head and a one-to-one prediction head.

[0081] In some embodiments, the training module 702 learns and trains the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model, including: dividing the satellite antenna image data set into a satellite antenna image data training set and a satellite antenna image data test set according to a preset ratio; training the constructed improved YOLO v10 network structure on the tensorfow system based on the satellite antenna image data training set to obtain an initial satellite antenna target detection model; the system parameters set by the tensorfow system include an initial learning rate and an iteration parameter of the initial learning rate; inputting the satellite antenna image in the satellite antenna image data test set into the initial satellite antenna target detection model to obtain a detection result, and comparing the detection result with the marked actual result, and verifying the detection accuracy of the initial satellite antenna target detection model according to the comparison result; if the detection accuracy of the initial satellite antenna target detection model meets the preset requirements, the initial satellite antenna target detection model is used as the trained satellite antenna target detection model.

[0082] The present application provides an antenna detection device based on improved YOLO v10, which obtains a pre-processed satellite antenna image dataset through an acquisition module; constructs an improved YOLO v10 network structure through a training module, and learns and trains the improved YOLO v10 network structure based on the satellite antenna image dataset to obtain a satellite antenna target detection model; obtains a satellite antenna image to be identified through a detection module, and inputs the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified. The antenna recognition method based on the single-stage detection YOLO algorithm can improve the detection speed of the satellite antenna; and in other embodiments, the data is enhanced using the Poisson fusion principle and the traditional image augmentation method, which can not only increase the number of data sets, but also improve the accuracy and generalization ability of the target detection model.

[0083] Based on the same concept of the present invention, the specification is attached Figure 8As shown, the structure of an electronic device 800 provided in an embodiment of the present application includes: at least one processor 801, at least one network interface 804 or other user interface 803, a memory 805, and at least one communication bus 802. The communication bus 802 is used to realize the connection and communication between these components. The electronic device 800 optionally includes a user interface 803, including a display (for example, a touch screen, LCD, CRT, holographic imaging (Holographic) or projection (Projector), etc.), a keyboard or a pointing device (for example, a mouse, a trackball (trackball), a touch pad or a touch screen, etc.).

[0084] The memory 805 may include a read-only memory and a random access memory, and provides instructions and data to the processor 801. A portion of the memory 805 may also include a non-volatile random access memory (NVRAM).

[0085] In some implementations, the memory 805 stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof:

[0086] Operating system 8051, including various system programs, used to implement various basic services and handle hardware-based tasks;

[0087] The application module 8052 includes various application programs, such as a launcher, a media player, a browser, etc., which are used to implement various application services.

[0088] In an embodiment of the present application, by calling the program or instruction stored in the memory 805, the processor 801 is used to execute steps in an antenna detection method based on an improved YOLO v10, and can complete the antenna identification of drone aerial photography based on the single-stage detection YOLO algorithm, which is more accurate and efficient.

[0089] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the antenna detection method based on the improved YOLO v10 are executed.

[0090] Specifically, the storage medium can be a general storage medium, such as a mobile disk, a hard disk, etc. When the computer program on the storage medium is run, the above-mentioned antenna detection method based on the improved YOLO v10 can be executed.

[0091] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of devices or units can be electrical, mechanical or other forms.

[0092] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0093] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0094] If the function is implemented in the form of a business function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0095] Finally, it should be noted that the above embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An antenna detection method based on improved YOLO v10, characterized in that: The method comprises the following steps: Acquire a preprocessed satellite antenna image dataset; wherein the preprocessed satellite antenna image dataset includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images; An improved YOLO v10 network structure is constructed, and the improved YOLOv10 network structure is trained based on the satellite antenna image data set to obtain a satellite antenna target detection model; wherein the improved YOLO v10 network structure replaces the PSA module in the original YOLO v10 network structure with a Double Attention module; The satellite antenna image to be identified is obtained, and the satellite antenna image to be identified is input into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

2. According to claim 1, an antenna detection method based on improved YOLO v10 is characterized in that: The step of obtaining the preprocessed satellite antenna image dataset comprises the following steps: Acquire a plurality of satellite antenna images, and pre-process the acquired plurality of satellite antenna images to filter out images that do not contain satellite antennas; Performing image augmentation processing on the plurality of pre-processed satellite antenna images to obtain a satellite antenna image set; wherein the image augmentation processing includes one or more of salt and pepper noise, Gaussian noise, Gaussian blur, brightness enhancement, and Poisson fusion; Data annotation is performed on each satellite antenna image in the obtained satellite antenna image set, and a satellite antenna image data set is obtained; wherein the data annotation includes the category of the satellite antenna and a bounding box in the image.

3. According to claim 2, an antenna detection method based on improved YOLO v10 is characterized in that: in, Use drones to take aerial photos of buildings and obtain multiple satellite antenna images.

4. According to claim 2, an antenna detection method based on improved YOLO v10 is characterized in that: Poisson fusion is performed in the following manner, including the following steps: Import the target satellite antenna image and background image; Replacing pixels at corresponding positions of the background image with pixels of the target satellite antenna image, and calculating the divergence of the target satellite antenna image; Iteratively solving the Poisson equation using the divergence and the values ​​of the pixels corresponding to the outermost part of the target satellite antenna image to obtain the pixel values ​​of the fused target satellite antenna image; The pixel values ​​of the target position in the background image are replaced by the pixel values ​​of the fused target satellite antenna image to obtain a final fused image.

5. According to claim 3, an antenna detection method based on improved YOLO v10 is characterized in that: The improved YOLO v10 network structure includes a backbone network, a neck and a prediction head, wherein the backbone network adopts CSPNet; the neck adopts a PAN structure; and the prediction head includes a one-to-many prediction head and a one-to-one prediction head.

6. The antenna detection method based on improved YOLO v10 according to claim 5, characterized in that: The step of learning and training the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model includes the following steps: Dividing the satellite antenna image data set into a satellite antenna image data training set and a satellite antenna image data test set according to a preset ratio; Training the constructed improved YOLO v10 network structure based on the satellite antenna image data training set to obtain an initial satellite antenna target detection model; Inputting the satellite antenna images in the satellite antenna image data test set into the initial satellite antenna target detection model to obtain a detection result, and comparing the detection result with the annotated actual result, and verifying the detection accuracy of the initial satellite antenna target detection model according to the comparison result; If the detection accuracy of the initial satellite antenna target detection model meets the preset requirements, the initial satellite antenna target detection model is used as the trained satellite antenna target detection model.

7. The antenna detection method based on improved YOLO v10 according to claim 6, characterized in that: in, The improved YOLO v10 network structure constructed is trained on the tensorfow system based on the satellite antenna image data training set to obtain an initial satellite antenna target detection model; the system parameters set by the tensorfow system include an initial learning rate and an iteration parameter of the initial learning rate.

8. An antenna detection device based on improved YOLO v10, characterized in that: The device comprises: An acquisition module, used to acquire a preprocessed satellite antenna image data set; wherein the preprocessed satellite antenna image data set includes a plurality of satellite antenna images and data annotations for each of the satellite antenna images; A training module is used to construct an improved YOLO v10 network structure, and to perform learning and training on the improved YOLO v10 network structure based on the satellite antenna image data set to obtain a satellite antenna target detection model; wherein the improved YOLO v10 network structure replaces the PSA module in the original YOLO v10 network structure with a Double Attention module; The detection module is used to obtain the satellite antenna image to be identified, and input the satellite antenna image to be identified into the satellite antenna target detection model to obtain the position and category of the satellite antenna in the satellite antenna image to be identified.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory through the bus, and when the machine-readable instructions are executed by the processor, the steps of the antenna detection method based on the improved YOLO v10 are performed as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the antenna detection method based on the improved YOLO v10 are executed.

Images