DnCNN and YOLOv8-based frequency height map denoising and identification method

Abstract

The invention provides a DnCNN and YOLOv8-based frequency height image denoising and recognition method. The method comprises the steps of inputting an ionized layer detection image into a DnCNN model to obtain a denoised clear image; and inputting the denoised clear image into a YOLOv8 model to obtain a recognition object and outputting ionosphere parameters. Through the precisely adjusted DnCNN model, noise in the ionized layer image is effectively removed, key echo information is reserved, and high-quality input is provided for subsequent image processing; through fine adjustment of the YOLOv8 model, resource consumption is reduced, and the efficiency of model training and the accuracy of object recognition are improved; automation and intellectualization of ionosphere detection image processing are realized, and processing efficiency and practicability are improved. The method is suitable for scientific research and monitoring application requiring high-precision ionosphere image analysis, and has important practical value and scientific significance.

Description

Technical Field

[0001] The present invention belongs to the field of high-frequency graph processing, and specifically relates to a high-frequency graph denoising and recognition method based on DnCNN and YOLOv8. Background Art

[0002] The ionosphere is a region in the Earth's atmosphere from about 60 km to 1000 km in altitude, where the air is ionized by solar radiation to form charged particles. This layer has an important influence on the propagation of radio waves and is an important channel for electromagnetic communication between the Earth and space. Disturbances in the ionosphere will directly affect the positioning accuracy of the global navigation satellite system and the stability and reliability of shortwave communications.

[0003] The frequency-height diagram is an important tool for studying the ionosphere. It is obtained through the ionospheric vertical altimeter. The horizontal and vertical coordinates represent the frequency and virtual height respectively, and the signal value of each coordinate point represents the echo intensity. The frequency-height diagram provides a basic description of ionospheric characteristics such as electron density and altitude, and shows the relationship between the height of the reflector and the frequency. However, there is a lot of noise in the frequency-height diagram, resulting in a low signal-to-noise ratio, which affects the accurate analysis and application of ionospheric characteristics.

[0004] Existing methods for processing frequency-height maps mainly include traditional image processing and deep learning methods. Traditional methods such as the ARTIST algorithm represent the electron density profile of each layer as a shifted Chebyshev polynomial. Although it has achieved certain results, it is difficult to adapt to the complexity and variability of the ionospheric structure. Other methods use digital image processing, mathematical morphology and graph theory techniques to measure frequency-height maps, but these methods rely on geographic location information, and the processing rules in different regions are quite different. Some technologies such as Steger curve tracking and statistical feature recognition, although not dependent on time, location and season, require complex adjustments when extended to other layers and are easily affected by noise. In general, traditional methods often rely on manual feature selection, are difficult to adapt to the complex structure of frequency-height maps, have weak generalization performance and are easily affected by noise. With the development of deep learning technology, a variety of image classification, semantic segmentation and target detection algorithms have been applied to frequency-height map measurement tasks. However, the performance of these methods is limited by the quantity and quality of labeled data, and it is difficult to obtain valuable results for images with excessive noise.

[0005] Common problems of traditional and deep learning methods include: difficulty in handling complex structures and diverse data sets, susceptibility to noise interference, reliance on manually selected features or specific statistical information, and weak generalization performance. Summary of the invention

[0006] The purpose of the present invention is to overcome the above problems and provide a frequency-height image denoising and recognition method based on DnCNN and YOLOv8. The present invention solves the problems of large noise interference and inaccurate key parameter extraction in the existing ionospheric detection image processing technology.

[0007] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0008] A frequency-high-frequency image denoising and recognition method based on DnCNN and YOLOv8, the method comprises the following steps:

[0009] The ionospheric detection image is input into the DnCNN model to obtain a denoised and clear image;

[0010] The training method of the DnCNN model includes the following steps: using a denoising data set to train a DnCNN basic model, using mean square error as a loss function;

[0011] Input the denoised clear image into the YOLOv8 model to obtain the identified object and output the ionospheric parameters;

[0012] The training method of the YOLOv8 model includes the following steps: pre-training a YOLOv8 basic model using natural images, and fine-tuning the pre-trained YOLOv8 basic model using a cross entropy loss function and a bounding box regression loss function.

[0013] As a further improvement, the construction of the denoising data set includes the following steps:

[0014] Collecting ionospheric sounding images at different geographical locations and under various climatic conditions, and dividing the ionospheric sounding images into a first group of ionospheric sounding images and a second group of ionospheric sounding images;

[0015] Performing artificial denoising and echo information retention processing on the first group of ionospheric sounding images to obtain a third group of ionospheric sounding images;

[0016] The second group of ionospheric sounding images and the third group of ionospheric sounding images are paired to construct a denoising data set for model supervised training; wherein the denoising data set is 80% of the training set and 20% of the validation set, which is used for model training and performance evaluation.

[0017] As a further improvement, the training method of the YOLOv8 model further includes the following steps:

[0018] The Adam optimizer and early stopping method are used to train the DnCNN model;

[0019] Among them, in the process of training the DnCNN model, when the loss on the validation set no longer decreases after multiple consecutive DnCNN models have completely traversed the data set, the training is stopped; among them, after each DnCNN model has completely traversed the data set, the denoising performance of the DnCNN model is evaluated on the validation set.

[0020] As a further improvement, the process of training the DnCNN model further includes the following steps:

[0021] Save the DnCNN model parameters that perform best on the validation set;

[0022] Load the best DnCNN model parameters on the validation set and evaluate it on the test set.

[0023] As a further improvement, after obtaining the denoised clear image, the denoised clear image needs to be annotated, including the following steps:

[0024] Annotate denoised clear image metrics;

[0025] The results of manual annotation measurement are paired with denoised clear images to create a target detection dataset for supervised training of the YOLOv8 model; wherein the target detection dataset is a 60% training set, a 20% validation set, and a 20% test set, which is used for model training and performance evaluation.

[0026] As a further improvement, the training method of the YOLOv8 model further includes the following steps:

[0027] During the fine-tuning of the YOLOv8 model, training is stopped when the loss on the validation set no longer decreases after the YOLOv8 model has completed multiple consecutive passes through the dataset. The metric accuracy of the YOLOv8 model is evaluated on the validation set after each pass through the dataset.

[0028] As a further improvement, the process of fine-tuning the YOLOv8 model further includes the following steps:

[0029] Save the YOLOv8 model parameters that perform best on the validation set;

[0030] Load the optimal YOLOv8 model parameters and evaluate it on the test set.

[0031] As a further improvement, the training method of the YOLOv8 model further includes the following steps:

[0032] Based on the characteristics of the ionospheric image, the output layer of the YOLOv8 model is adjusted to align the number of categories in the classification head with the number of categories in the layers of the ionospheric image.

[0033] As a further improvement, the method further comprises the following steps:

[0034] Convert the ionospheric parameters of the YOLOv8 model into critical frequency and virtual height;

[0035] The formulas for the conversion critical frequency and virtual height are as follows:

[0036]

[0037] Where f is the critical frequency, h is the virtual height, δf is the unit measurement of the horizontal axis of the ionospheric frequency-height diagram, δh is the unit measurement of the vertical axis of the ionospheric frequency-height diagram, and f is 0 Indicates the starting frequency of the horizontal and vertical axes of the ionospheric frequency diagram, h 0 Indicates the starting height of the horizontal and vertical axes of the ionospheric frequency diagram. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. represents the width of the bounding box, Indicates the height of the bounding box.

[0038] On the other hand, the present invention also provides a high-frequency graph denoising and recognition system based on DnCNN and YOLOv8, wherein the system comprises at least one processor and a memory storing instructions, and when the instructions are executed by at least one processor, the method in the above technical solution is implemented.

[0039] A computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method described in the steps of the above technical solution.

[0040] A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method described in the steps of the above technical solution is implemented.

[0041] The beneficial effects of the present invention are as follows: through the precisely adjusted DnCNN model, the noise in the ionospheric image is effectively removed, the key echo information is retained, and high-quality input is provided for subsequent image processing; through the fine-tuning of the YOLOv8 model, the resource consumption is reduced, the efficiency of model training and the accuracy of object recognition are improved; the automation and intelligence of ionospheric detection image processing are realized, and the processing efficiency and practicality are improved; it is suitable for scientific research and monitoring applications that require high-precision ionospheric image analysis, and has important practical value and scientific significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention has the following accompanying drawings:

[0043] Figure 1 It is a flowchart of a frequency-high image denoising and recognition method based on DnCNN and YOLOv8 in some specific embodiments of the present invention;

[0044] Figure 2 is a manual denoising flow chart of some specific embodiments of the present invention;

[0045] Figure 3 is a flow chart of manual measurement of some specific embodiments of the present invention;

[0046] Figure 4 A schematic diagram of a system for denoising and identifying high-frequency graphs based on DnCNN and YOLOv8 in the present invention. DETAILED DESCRIPTION

[0047] The following examples further illustrate the content of the present invention, but should not be construed as limiting the present invention. Without departing from the spirit and substance of the present invention, modifications or substitutions made to the methods, steps or conditions of the present invention all fall within the scope of the present invention.

[0048] Embodiment 1:

[0049] like Figure 1-3 As shown, the present invention provides a frequency map denoising and recognition method based on DnCNN and YOLOv8, the method comprising the following steps:

[0050] Step S101: input the ionospheric detection image into the DnCNN model to obtain a denoised and clear image;

[0051] The training method of the DnCNN model includes the following steps: using a denoising data set to train a DnCNN basic model, using mean square error as a loss function;

[0052] Step S102: inputting the denoised clear image into the YOLOv8 model to obtain the identified object and output ionospheric parameters;

[0053] For example, when a denoised and clear image is input into a fine-tuned YOLOv8 model, the model not only recognizes the target object in the image, but also accurately measures and outputs the required ionospheric parameters.

[0054] The training method of the YOLOv8 model includes the following steps: pre-training a YOLOv8 basic model using natural images, and fine-tuning the pre-trained YOLOv8 basic model using a cross entropy loss function and a bounding box regression loss function.

[0055] Among them, the formula of the loss function is:

[0056] is the total loss, is the classification loss, is the positioning loss, is the target loss; CIOU ,λ obj ,λ cls It is a weight coefficient used to balance the contribution of each part of the loss function.

[0057] b is the true bounding box, predicted bounding box CIOU: represents the complete intersection-over-union ratio between the predicted bounding box and the true bounding box.

[0058] y represents the true label, which is 1 if there is a target, otherwise it is 0; It indicates the model predicts the probability that the current box contains the target.

[0059] C is the total number of categories; if the true category of the object is c, it is 1, otherwise it is 0; The model predicts the probability that the current box belongs to category c.

[0060] For example, fine-tuning the pre-trained YOLOv8 model: Use the YOLOv8 model pre-trained on natural images and fine-tune it based on the characteristics of ionospheric images; when fine-tuning, use the cross entropy loss function and bounding box regression loss to optimize the model to ensure that the model can accurately identify and locate key features in ionospheric images.

[0061] The construction of the denoising dataset includes the following steps:

[0062] Step S1011: collecting ionospheric sounding images at different geographical locations and under various climatic conditions, and dividing the ionospheric sounding images into a first group of ionospheric sounding images and a second group of ionospheric sounding images;

[0063] For example, collecting and selecting frequency-height maps: collecting ionospheric sounding images from different geographical locations and various climatic conditions, and selecting some images as the basis for the denoising dataset.

[0064] Step S1012: performing artificial denoising and echo information retention processing on the first group of ionospheric sounding images to obtain a third group of ionospheric sounding images;

[0065] For example, manual processing and annotation: the selected images are manually processed, including removing noise and retaining echo information; this ensures that the model can focus on noise features in subsequent training.

[0066] Step S1013: Pair the second group of ionospheric detection images with the third group of ionospheric detection images to construct a denoised data set for model supervised training; wherein the denoised data set is 80% of the training set and 20% of the validation set, which are used for model training and performance evaluation.

[0067] For example, to establish training and validation datasets: pair the manually processed clean images with the original noisy images to create a denoising dataset for supervised model training; the dataset is divided into 80% training set and 20% validation set for model training and performance evaluation.

[0068] The training method of the YOLOv8 model also includes the following steps:

[0069] The Adam optimizer and early stopping method are used to train the DnCNN model;

[0070] Among them, in the process of training the DnCNN model, when the loss on the validation set no longer decreases after multiple consecutive DnCNN models have completely traversed the data set, the training is stopped; among them, after each DnCNN model has completely traversed the data set, the denoising performance of the DnCNN model is evaluated on the validation set.

[0071] For example, the Adam optimizer is used for model training, the learning rate is set to 0.001, and batches with a batch size of 32 are used for iteration; the early stopping method is used during training to avoid overfitting, and the model performance is regularly evaluated on the validation set to adjust the learning rate.

[0072] The process of training the DnCNN model also includes the following steps:

[0073] During the training of the DnCNN model, the DnCNN model parameters with the best performance on the validation set are saved;

[0074] Load the best DnCNN model parameters on the validation set and evaluate it on the test set.

[0075] After obtaining the denoised clear image, it is necessary to annotate the denoised clear image, which includes the following steps:

[0076] Annotate denoised clear image metrics;

[0077] For example, annotating denoised images: further annotating all output denoised images to identify key objects in the image and their corresponding ionospheric parameters, such as false height and critical frequency.

[0078] The results after manual annotation measurement are paired with denoised clear images (referring to the images output by DnCNN. Since DnCNN is a denoising model, the images it outputs are called denoised clear images. The denoised clear images after annotation are paired with the denoised clear images before annotation) to create a target detection dataset for supervised training of the YOLOv8 model; wherein the target detection dataset is a 60% training set, a 20% validation set, and a 20% test set, which is used for model training and performance evaluation.

[0079] The training method of the YOLOv8 model also includes the following steps:

[0080] During the fine-tuning of the YOLOv8 model, training is stopped when the loss on the validation set no longer decreases after the YOLOv8 model has completed multiple consecutive passes through the dataset. The metric accuracy of the YOLOv8 model is evaluated on the validation set after each pass through the dataset.

[0081] The process of fine-tuning the YOLOv8 model also includes the following steps:

[0082] When fine-tuning the YOLOv8 model, save the YOLOv8 model parameters that perform best on the validation set;

[0083] Load the optimal YOLOv8 model parameters and evaluate it on the test set.

[0084] The training method of the YOLOv8 model further includes the following steps:

[0085] Based on the characteristics of the ionospheric image, the output layer of the YOLOv8 model is adjusted to align the number of types in the classification head with the number of types in the ionospheric image. The output of the YOLOv8 model is

[0086]

[0087] In the formula, Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. Indicates the vertical coordinate of the center of the bounding box relative to the upper left corner of the image. Indicates the width of the bounding box. Indicates the height of the bounding box. represents the target confidence, represents the classification confidence of each category, and C represents the total number of categories.

[0088] In some embodiments of the present invention, the measurement method in the frequency-frequency graph denoising and recognition method based on DnCNN and YOLOv8 further includes the following steps:

[0089] Convert the ionospheric parameters of the YOLOv8 model into critical frequency and virtual height;

[0090] The formulas for the conversion critical frequency and virtual height are as follows:

[0091]

[0092] Where f is the critical frequency, h is the virtual height, δf is the unit measurement of the horizontal axis of the ionospheric frequency-height diagram, δh is the unit measurement of the vertical axis of the ionospheric frequency-height diagram, and f is 0 Indicates the starting frequency of the horizontal and vertical axes of the ionospheric frequency diagram, h 0 Indicates the starting height of the horizontal and vertical axes of the ionospheric frequency diagram. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. represents the width of the bounding box, Indicates the height of the bounding box.

[0093] Embodiment 2:

[0094] The present invention provides a frequency-high-frequency graph denoising and recognition method based on a DnCNN model and a YOLOv8 model, the method comprising the following steps:

[0095] Step S1: Input the original frequency-height map into the pre-established and trained DnCNN model. The DnCNN model learns the noise characteristics of the frequency-height map through multiple convolutional layers, performs denoising, and outputs a high-quality denoised frequency-height map;

[0096] Step S2: Use the YOLOv8 model to perform target detection on the denoised frequency-height map;

[0097] Step S3: The YOLOv8 model uses the convolutional layer to extract features and locate targets on the frequency map;

[0098] Step S4: Use the YOLOv8 model to extract different layer parameters in the frequency height map, including false height and critical frequency. The model effectively identifies the key features and parameters in the frequency height map through precise bounding box positioning.

[0099] In the present invention, a method combining DnCNN and YOLOv8 models is proposed, and DnCNN is used to perform frequency height map denoising to improve the quality and signal-to-noise ratio of the image; then the YOLOv8 model is used to perform target detection and parameter extraction to improve the accuracy and efficiency of detection. The YOLOv8 model can accurately identify and extract different layer parameters in the frequency height map through precise bounding box positioning, and is suitable for the analysis and application of complex frequency height maps; the method of the present invention is suitable for a variety of ionospheric monitoring scenarios, and has broad application prospects and practical value; through the method of the present invention, the signal-to-noise ratio and detection accuracy of the frequency height map can be effectively improved, providing reliable technical support for ionospheric monitoring and communication system optimization.

[0100] Embodiment 3:

[0101] The present invention provides a method for denoising, identifying and measuring a high-frequency graph based on a DnCNN model and a YOLOv8 model. The specific flow chart is as follows: Figure 1 As shown, the training process of the DnCNN denoising model and the YOLOv8 target detection model and the use of these models for frequency map denoising and target detection tasks are shown below:

[0102] 1. The training process of the DnCNN denoising model includes the following steps:

[0103] Step S11: Collect ionospheric frequency and height maps obtained from different geographical locations around the world under various climatic conditions to ensure the diversity and extensiveness of the data.

[0104] Step S12: Select representative images from the collected data, especially those with high noise levels, for training the denoising model.

[0105] Step S13: artificially adding different types of noise (such as Gaussian noise) to the selected image to form more noisy images.

[0106] Step S14: Perform artificial denoising on the selected image to fill in the breakpoints of the discontinuous signal and retain the key echo information for supervising the training of the denoising model DnCNN. For specific denoising examples, refer to Figure 2 .

[0107] Step S15: Establish training and verification data sets. Pair the artificially denoised image with the original noisy image (I, I denoise ), create a denoising dataset for supervised training of the DnCNN model. The dataset is divided into 60% training set, 20% validation set and 20% test set for model training and performance evaluation.

[0108] Step S16: Select the DnCNN model as the denoising basis, which effectively learns the noise distribution and performs denoising through a deep convolutional network. The output of the convolutional layer is

[0109] h l =ReLU(BN(W l *h l-1 +b l )): ReLU(x)=max(0,x), h l It is the first

[0110] The output of the convolution layer, W l is the convolution kernel of the lth layer, b l is the bias of the lth layer of convolution, * is the convolution operation; μ and σ 2 are the mean and variance of the batch, and γ and β are the scaling and offset factors.

[0111] Step S17: Using mean square error (MSE) as the loss function

[0112] N is the batch size, y i is the original clean image, f(x i ; θ) is the image after model denoising.

[0113] Directly measures the difference between the denoised image and the original clean image.

[0114] Step S18: Use the Adam optimizer to train the model. Its adaptive learning rate can accelerate the convergence speed and improve the denoising effect. The specific method of the Adam optimizer is: First, θ 0 =0, m 0 =0, v 0 =0,θ 0 is the initialization parameter of the model, m 0 and v 0 is the first-order moment and second-order moment estimate of the initial gradient; then, the gradient of the current loss function is calculated and the moment estimate is updated, β 1 , β 2 is the decay rate of the first-order and second-order moment estimates; the moment estimates are corrected, Use the moment estimate of the gradient to update the model parameters α is the learning rate, ∈ avoids the denominator being zero.

[0115] Step S19: Implement early stopping during training to prevent overfitting, and stop training when the loss on the validation set no longer decreases for multiple epochs (the number of times the model fully traverses the data set).

[0116] Step S20: After each training epoch (the number of times the model completely traverses the data set), the denoising performance of the model is evaluated on the validation set to ensure the generalization ability of the model, and the model parameters that perform best on the validation set during the training process are saved.

[0117] Step S21: Load the best model parameters on the validation set and evaluate them on the test set.

[0118] Step S22: input all the original ionospheric frequency-height maps in step S11 into the model to obtain high-quality automatically denoised frequency-height maps corresponding to all the original ionospheric frequency-height maps.

[0119] 2. Training and fine-tuning of the YOLOv8 model target detection model:

[0120] Step S23: further annotate and measure all denoised images output from step S22 to identify key objects in the image and their corresponding ionospheric parameters, such as false height and critical frequency. For specific annotation examples, refer to Figure 3 .

[0121] Step S24: Create training and validation datasets. Pair the results of manual annotation measurements with the denoised clear images to create an object detection dataset for supervised training of the YOLOv8 model. The dataset is divided into a 60% training set, a 20% validation set, and a 20% test set for model training and performance evaluation.

[0122] Step S25: Download the YOLOv8 model.

[0123] Step S26: According to the characteristics of the ionospheric image, the output layer of the YOLOv8 model is adjusted to align the number of categories in the classification head with the number of categories in the layer of the ionospheric image.

[0124] Output of the YOLOv8 model

[0125]

[0126] in: is the coordinate of the center of the bounding box relative to the upper left corner of the image;

[0127] are the width and height of the bounding box;

[0128] is the target confidence;

[0129] is the classification confidence of each category;

[0130] C is the total number of categories.

[0131] Step S27: Download and load the pre-trained weights on the COCO natural image dataset (natural images refer to images that actually exist in nature. These images are produced by photographing natural processes or phenomena, rather than being created by artificial or technical means), and use its complex visual features learned on a wide range of datasets to provide a strong starting point for the specific task of ionospheric images.

[0132] Step S28: Fine-tune the pre-trained YOLOv8 model and use the cross entropy loss function and bounding box regression loss to optimize the model. The cross entropy loss function and bounding box regression loss function are as follows:

[0133] is the total loss, is the classification loss, is the positioning loss, is the target loss; CIOU ,λ obj ,λ cls It is a weight coefficient used to balance the contribution of each part of the loss function.

[0134] b is the true bounding box, predicted bounding box CIOU: represents the complete intersection-over-union ratio between the predicted bounding box and the true bounding box.

[0135] y represents the true label, which is 1 if there is a target, otherwise it is 0; It indicates the model predicts the probability that the current box contains the target.

[0136] C is the total number of categories; if the true category of the object is c, it is 1, otherwise it is 0; The model predicts the probability that the current box belongs to category c.

[0137] Step S29: Implement early stopping during fine-tuning to prevent overfitting, and stop training when the loss on the validation set no longer decreases for multiple epochs (the number of times the model fully traverses the dataset).

[0138] Step S30: After each training epoch (the number of times the model completely traverses the data set), the metric accuracy of the YOLOv8 model is evaluated on the validation set to ensure the generalization ability of the model, and the model parameters that perform best on the validation set during the fine-tuning process are saved.

[0139] Step S31: Load the optimal model parameters and evaluate them on the test set.

[0140] Step S32: Use the SGD optimizer with momentum for training, and set the initial learning rate to 0.0001 and the momentum factor to 0.9. The specific method of the SGD optimizer is: First, calculate the parameter θ t-1 The loss function at (parameter value at the previous moment) Gradient Then, using the momentum V from the previous time step t-1 and the current gradient g t , update the momentum V according to the momentum factor γ and learning rate α t =γV t-1 +αg t ; Finally, use the updated momentum V t Update the parameter θ t =θ t-1 -V t .

[0141] Step S33: input all ionospheric frequency-height maps that have been denoised in step S22 into the model to obtain automatic measurement results of all ionospheric frequency-height maps.

[0142] Step S34: Through mathematical calculations, the target detection results of YOLOv8 are converted into numerical results with physical meaning.

[0143]

[0144] Among them, f is the critical frequency, h is the virtual height, δfδh are the unit measurements of the horizontal and vertical axes of the ionospheric frequency-height diagram, respectively, and f 0 h 0 They are the starting frequency and starting height of the horizontal and vertical axes of the ionospheric frequency-height diagram, respectively. The calculation parameters are obtained as a reference Figure 2 .

[0145] Through the precisely adjusted DnCNN model, the noise in the ionospheric image is effectively removed, the key echo information is retained, and high-quality input is provided for subsequent image processing; through fine-tuning the YOLOv8 model, resource consumption is reduced, the efficiency of model training and the accuracy of object recognition are improved; the automation and intelligence of ionospheric detection image processing are realized, and the processing efficiency and practicality are improved; it is suitable for scientific research and monitoring applications that require high-precision ionospheric image analysis, and has important practical value and scientific significance.

[0146] In some specific embodiments, Figure 4As shown, a system for denoising and identifying high-frequency graphs based on the DnCNN model and the YOLOv8 model includes at least one processor and a memory storing instructions to implement any of the high-frequency graph denoising and identification methods based on DnCNN and YOLOv8 in the above technical solutions.

[0147] In some specific embodiments, a computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps in the above technical solution.

[0148] In some specific embodiments, a computer program product is provided, the computer program product includes a computer program, and when the computer program is executed by a processor, the method of implementing the steps in the above technical solution is implemented.

[0149] The embodiments and functional operations of the subject matter described in this specification may be implemented in: digital electronic circuitry, tangibly implemented computer software or firmware, computer hardware, including the structures disclosed in this specification and their structural equivalents, or a combination of more than one of the above. The embodiments of the subject matter described in this specification may be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on one or more tangible non-transitory program carriers, for execution by a data processing device or to control the operation of a data processing device.

[0150] Alternatively or additionally, the program instructions may be encoded on an artificially generated propagation signal, such as a machine-generated electrical, optical, or electromagnetic signal, which is generated to encode information for transmission to an appropriate receiver device for execution by a data processing device. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or a combination of one or more of the above devices.

[0151] A computer program (which may also be referred to or described as a program, software, software application, module, software module, script or code) may be written in any form of programming language, including compiled or interpreted languages ​​or declarative or procedural languages, and the computer program may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a file in a file system. A program may be stored in a portion of a file that stores other programs or data, for example, one or more scripts stored in the following: in a markup language document; in a single file dedicated to a related program; or in multiple collaborative files, for example, files storing one or more modules, subroutines or code portions. A computer program may be deployed to execute on one or more computers, the computers being located in one place or distributed to multiple locations and interconnected by a communication network.

[0152] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

[0153] Although the present invention has been described in detail above by general description, specific implementation methods and experiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. Frequency map denoising and recognition method based on DnCNN and YOLOv8, It is characterized in that The following steps are involved: The ionospheric detection image is input into the DnCNN model to obtain a denoised and clear image; The training method of the DnCNN model includes the following steps: using a denoising data set to train a DnCNN basic model, using mean square error as a loss function; Input the denoised clear image into the YOLOv8 model to obtain the identified object and output the ionospheric parameters; The training method of the YOLOv8 model includes the following steps: pre-training a YOLOv8 basic model using natural images, and fine-tuning the pre-trained YOLOv8 basic model using a cross entropy loss function and a bounding box regression loss function.

2. According to the method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 in claim 1, It is characterized in that The construction of the denoising dataset includes the following steps: Collecting ionospheric sounding images at different geographical locations and under various climatic conditions, and dividing the ionospheric sounding images into a first group of ionospheric sounding images and a second group of ionospheric sounding images; Performing artificial denoising and echo information retention processing on the first group of ionospheric sounding images to obtain a third group of ionospheric sounding images; The second group of ionospheric sounding images and the third group of ionospheric sounding images are paired to construct a denoising data set for model supervised training; wherein the denoising data set is 80% of the training set and 20% of the validation set, which is used for model training and performance evaluation.

3. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 1, It is characterized in that The training method of the YOLOv8 model further includes the following steps: The Adam optimizer and early stopping method are used to train the DnCNN model; Among them, in the process of training the DnCNN model, when the loss on the validation set no longer decreases after multiple consecutive DnCNN models have completely traversed the data set, the training is stopped; among them, after each DnCNN model has completely traversed the data set, the denoising performance of the DnCNN model is evaluated on the validation set.

4. According to the method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 in claim 3, It is characterized in that The process of training the DnCNN model also includes the following steps: Save the DnCNN model parameters that perform best on the validation set; Load the best DnCNN model parameters on the validation set and evaluate it on the test set.

5. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 1, It is characterized in that After obtaining the denoised clear image, the denoised clear image needs to be annotated, including the following steps: Annotate denoised clear image metrics; The results of manual annotation measurement are paired with denoised clear images to create a target detection dataset for supervised training of the YOLOv8 model; wherein the target detection dataset is a 60% training set, a 20% validation set, and a 20% test set, which is used for model training and performance evaluation.

6. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 1, It is characterized in that The training method of the YOLOv8 model further includes the following steps: During the fine-tuning of the YOLOv8 model, training is stopped when the loss on the validation set no longer decreases after the YOLOv8 model has completed multiple consecutive passes through the dataset. The metric accuracy of the YOLOv8 model is evaluated on the validation set after each pass through the dataset.

7. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 6, It is characterized in that The process of fine-tuning the YOLOv8 model also includes the following steps: Save the YOLOv8 model parameters that perform best on the validation set; Load the optimal YOLOv8 model parameters and evaluate it on the test set.

8. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 1, It is characterized in that The training method of the YOLOv8 model further includes the following steps: Based on the characteristics of the ionospheric image, the output layer of the YOLOv8 model is adjusted to align the number of categories in the classification head with the number of categories in the layers of the ionospheric image.

9. The method for denoising and identifying high-frequency images based on DnCNN and YOLOv8 according to claim 1, It is characterized in that It further includes the following steps: Convert the ionospheric parameters of the YOLOv8 model into critical frequency and virtual height; The formulas for the conversion critical frequency and virtual height are as follows: Where f is the critical frequency, h is the virtual height, δf is the unit measurement of the horizontal axis of the ionospheric frequency-height diagram, δh is the unit measurement of the vertical axis of the ionospheric frequency-height diagram, and f is 0 Indicates the starting frequency of the horizontal and vertical axes of the ionospheric frequency diagram, h 0 Indicates the starting height of the horizontal and vertical axes of the ionospheric frequency diagram. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. Indicates the horizontal coordinate of the center of the bounding box relative to the upper left corner of the image. represents the width of the bounding box, Indicates the height of the bounding box.

10. A frequency-height map denoising and recognition system based on DnCNN and YOLOv8, the system comprising at least one processor and a memory storing instructions, which, when executed by at least one processor, implements a geological acquisition terminal and a geological big data warehouse storage method as described in any one of claims 1 to 9.

Images