Airport pavement area snow measurement method based on GB-ArcSAR

Abstract

The invention relates to the technical field of snowfall monitoring, and discloses a GB-ArcSAR-based airport pavement area snow measurement method, which comprises the following steps of: generating an echo signal by using a GB-ArcSAR, optimizing signal sampling through a self-adaptive sampling technology, and reducing redundant calculation; performing multi-resolution analysis on the echo signal in combination with wavelet transform, and extracting snow layer features; the extraction of snow layer features is optimized through a convolutional neural network, and a snow accumulation area is automatically identified; adopting fast Fourier transform to accelerate frequency domain processing of the echo signal; the data processing efficiency is improved through parallel computing and a distributed computing framework, it is ensured that the system can feed back accumulated snow information in real time, and airport snow removal work and flight scheduling are supported. Through the technical means of adaptive sampling, deep learning optimization, fast Fourier transform acceleration, parallel calculation and the like, the efficiency and precision of accumulated snow amount measurement are remarkably improved, redundancy calculation and processing bottlenecks are avoided, an efficient and stable accumulated snow monitoring solution is provided, and efficient operation of an airport and flight safety are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of snowfall monitoring, and specifically to a method for measuring snow accumulation on airport runways based on GB-ArcSAR. Background Art

[0002] Snow accumulation monitoring on airport runways is crucial for flight safety and airport operations. Currently, common snow measurement methods include traditional radar technology, manual detection, and meteorological data extrapolation. These methods rely on radar echo signals or meteorological data, but under complex weather conditions, their accuracy and real-time performance often fail to meet requirements.

[0003] Traditional radar technology shows certain limitations in snow layer monitoring. Since radar echo signals are affected by various meteorological factors, especially when the snow layer is thin or weather conditions are complex, the accuracy of echo signals is relatively low, often leading to misjudgment of the snow layer. In addition, during the process of snow accumulation identification using traditional radar technology, it is difficult to effectively distinguish the reflection signals of snow accumulation from those of other surface substances.

[0004] Existing snowfall monitoring methods usually adopt a uniform sampling strategy, but this method will conduct a large number of redundant samplings on areas without snow accumulation, wasting computing resources and having low computing efficiency. Especially when the snow accumulation amount is small or the snow accumulation is uneven, the response speed and computing power of the system are greatly limited, making it difficult to achieve real-time and accurate monitoring.

[0005] Although advanced technologies such as deep learning have been applied in some fields, in snow accumulation monitoring, existing image processing technologies still fail to effectively solve the problem of complex feature extraction in snow accumulation echo signals. Traditional feature extraction methods rely on manually set rules or thresholds, lack flexibility, and cannot handle the variability of snow layers under different meteorological conditions, resulting in low detection accuracy.

[0006] Existing snow accumulation monitoring systems also have significant bottlenecks in large-scale data processing and real-time feedback. Most systems rely on a single computing unit, and when processing a large amount of data, the computing speed and real-time feedback ability are limited, which affects the efficiency of airport snow removal operations and the timeliness of flight scheduling.

[0007] Therefore, existing technologies are difficult to meet the requirements of high efficiency, accuracy, and real-time performance for measuring snow accumulation on airport runways, and there is an urgent need for a new technical solution that can overcome these deficiencies. Summary of the Invention

[0008] Aiming at the deficiencies of the existing technology, the present invention provides a method for measuring snow accumulation on airport runways based on GB-ArcSAR, which solves the problem that the existing airport runway snow accumulation measurement technology cannot efficiently, accurately, and real-time detect the snow accumulation amount under complex meteorological conditions.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solutions: A method for measuring snow accumulation on airport runways based on GB-ArcSAR, comprising the following steps: Use GB-ArcSAR to generate echo signals of the airport runway surface; Analyze the echo signals, establish a snow accumulation echo signal model, and deduce the electromagnetic reflection characteristics of the snow accumulation; Based on the reflection intensity of the echo signals, adopt an adaptive sampling method to adjust the sampling density, focus on sampling the snow accumulation area, and reduce the calculation of the snow-free area; Perform multi-resolution analysis on the echo signals using wavelet transform, extract the main features of the signals and remove noise, reducing data redundancy; Use a convolutional neural network to extract features and classify the snow accumulation echo signals to identify the snow layer area; Adopt fast Fourier transform to accelerate the processing of echo signals and improve the calculation efficiency through frequency domain conversion; Combine the echo signal data with weather data, calculate the snow accumulation amount on the airport runway surface, and provide real-time feedback.

[0010] Preferably, the adaptive sampling method includes: Set the sampling density of each sampling point according to the intensity of the echo signals where A(x, y) is the intensity of the echo signal at position (x, y), and γ is a weighting coefficient that controls the relationship between the sampling density and the intensity of the echo signal; The value of the weighting coefficient γ is dynamically adjusted according to the variation range and noise level of the echo signals.

[0011] Preferably, the steps of the wavelet transform include: Perform wavelet transform on the echo signal x(t) to obtain wavelet transform coefficients at multiple scales where W(a, b) is the wavelet transform coefficient at scale a and displacement b, ψ(t) is the mother wavelet, x(t) is the echo signal, and the mother wavelet is the Daubechies wavelet, which has optimal time-frequency localization characteristics and is suitable for processing echo signals.

[0012] Preferably, the convolutional neural network includes the following steps: Use training sample data to train the network and learn the echo signal characteristics of the snow accumulation and non-snow accumulation areas; Extract the snow layer features in the echo signals through convolutional layers and pooling layers, and output the snow layer classification results; The convolutional neural network includes two convolutional layers and two pooling layers, where the convolutional kernel size of the convolutional layers is 3×3, and the pooling layers adopt 2×2 maximum pooling operations.

[0013] Preferably, the loss function L of the convolutional neural network is as follows: where y i is the actual label, is the predicted label of the convolutional neural network, and N is the number of training samples.

[0014] Preferably, the fast Fourier transform includes: Performing frequency-domain conversion on the echo signal, using the Fourier transform algorithm to convert the time-domain signal into a frequency-domain signal, thereby accelerating the processing of the echo signal and reducing the computational complexity; The calculation formula of the fast Fourier transform is: where X(k) is the frequency-domain signal, x(n) is the time-domain signal, N is the signal length, and k is the frequency index.

[0015] Preferably, the fast Fourier transform is used to accelerate the spectral analysis of the echo signal, especially to remove high-frequency noise and enhance the snow layer features in the signal.

[0016] Preferably, the method further includes the following steps: Using a parallel computing method, dividing the snow echo signal data into multiple subtasks and allocating them to a multi-core processor or a distributed computing environment for parallel processing to further improve the processing speed; The parallel computing method includes using the MapReduce computing framework to perform distributed processing on the echo signal data to accelerate the snow volume calculation process.

[0017] Preferably, the parallel computing method adopts multi-threaded processing to ensure real-time snow volume detection when dealing with a large amount of data.

[0018] Preferably, the echo signal data is combined with weather data, and the snow accumulation amount on the airport runway is calculated through a model. The weather data includes information such as temperature, humidity, and snowfall amount. By combining with the echo signal data, the snow accumulation amount can be accurately calculated.

[0019] The present invention provides an airport runway snow accumulation measurement method based on GB-ArcSAR. It has the following beneficial effects: 1. The present invention adopts an imaging technology based on GB-ArcSAR and an adaptive sampling method. By efficiently analyzing the snow echo signal, it realizes real-time monitoring of the snow accumulation amount on the airport runway. Compared with the traditional radar echo analysis method in the prior art, the present invention greatly improves the calculation efficiency, avoids redundant sampling in snow-free areas, thereby reducing the waste of computing resources, and effectively improves the real-time performance and processing ability of the system.

[0020] 2. By introducing deep learning to optimize snow layer feature extraction, the present invention not only automatically identifies snow-covered areas but also significantly improves the accuracy of snow depth measurement. Compared with traditional manual or rule-based methods, the present invention can process more complex snow echo signals, automatically learn the characteristics of snow under different meteorological conditions, and solve the deficiency that existing methods cannot effectively adapt to complex snow changes.

[0021] 3. The present invention performs frequency-domain processing on echo signals through fast Fourier transform, making signal analysis more efficient and accurate. Different from the traditional method that requires a large amount of time-domain calculations, FFT can quickly extract snow layer features and remove noise in the frequency domain, significantly accelerating the signal processing speed and improving the response ability of the snow monitoring system under complex weather conditions.

[0022] 4. The present invention combines parallel computing and distributed computing frameworks to effectively solve the bottleneck of large-scale data processing in snow monitoring. Compared with the single-machine processing method in the prior art, the present invention distributes tasks to multiple computing nodes through distributed computing, significantly improving the processing speed and ensuring that the system still maintains an efficient and stable working state in a large-scale data environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic flowchart of the steps of the method for measuring snow accumulation on the airport runway in the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0025] During the operation of the airport, snow accumulation is a factor directly affecting flight safety. Traditional snow measurement methods, such as manual measurement or detection based on traditional radars, have many limitations, such as low calculation efficiency, poor accuracy, and inability to provide effective data in real time under complex meteorological conditions. To address these problems, the present invention proposes a method for measuring snow accumulation on the airport runway based on GB-ArcSAR (Geometrically Imitative Reflector Synthetic Aperture Radar), aiming to achieve efficient and accurate snow depth measurement by optimizing the processing process of radar echo signals, especially under complex weather conditions.

[0026] GB-ArcSAR is a radar imaging technology capable of acquiring high-precision images. It obtains images of distant targets through synthetic aperture. Different from traditional radar imaging methods, GB-ArcSAR can provide clearer images and is especially suitable for working under complex surface conditions such as snow cover. By accurately analyzing the snow echo signal, key parameters such as snow thickness and humidity can be measured, thus accurately evaluating the snow accumulation on the airport runway.

[0027] The core principle of the present invention is to utilize the GB-ArcSAR imaging technology. By optimizing algorithms, especially reducing redundant calculations during the imaging process, the processing speed is accelerated to achieve real-time and accurate snow accumulation detection.

[0028] Please refer to the attached Figure 1 , the method for measuring snow accumulation on the airport runway includes the following steps: S1. Echo signal generation and analysis First of all, the generation and analysis of echo signals is the first step of the present invention and the basis of the entire snow accumulation measurement process. This step mainly involves using GB-ArcSAR technology to obtain the echo signals of the airport runway and, through the analysis of these signals, deriving the electromagnetic reflection characteristics of the snow. The echo signals contain information about snow thickness, humidity, and density, etc. These information provide basic data for subsequent snow volume calculation and snow layer feature extraction.

[0029] In this step, first, the airport runway is irradiated with radar waves, and the radar waves are reflected by the snow layer on the runway. The reflected signal is received by the radar and processed to obtain the corresponding echo signal. Through further analysis of these echo signals, the electromagnetic characteristics of the snow can be derived and data basis can be provided for subsequent signal processing.

[0030] In the specific implementation process, the generated echo signals usually undergo multiple reflections, especially when there are multiple different physical layers (such as the snow layer and the ground layer). The propagation characteristics of radar waves are affected by these layers, so it is necessary to accurately model the echo signals to ensure the accuracy of the reflection data.

[0031] In this embodiment, first, the echo signals are subjected to preliminary geometric correction and denoising processing to ensure that the signal quality meets the requirements of subsequent analysis. By geometrically modeling the echo signals, the propagation path of the radar waves is determined, the propagation distances of the reflected waves and the initial transmitted waves are calculated, and the reflection characteristics of different snow layers are analyzed. Specifically, the intensity and phase of the echo signals are affected by factors such as the thickness, humidity, density of the snow layer and the incident angle of the radar.

[0032] In some embodiments, the reflection characteristics of the echo signals can be described by the Fresnel equation, and the reflection coefficient R can be expressed as: Wherein: n 1 is the refractive index of air, n 2 is the refractive index of the snow cover layer, θ is the incident angle.

[0033] This formula can be used to calculate the reflection coefficient between the radar wave and the snow cover layer, thereby providing accurate data support for subsequent calculations of the snow depth, humidity, etc.

[0034] During the implementation process, it is also necessary to consider the influence of the thickness of the snow cover layer on the echo signal. Specifically, the thicker the snow cover layer, the more obvious the attenuation effect of the radar wave, so the intensity of the echo signal will also change. For this reason, the reflection coefficient and the radar wave propagation model are used to estimate the variation law of the echo signal under different snow layer thicknesses.

[0035] In addition, when analyzing the echo signal, it is also necessary to consider the propagation time of the radar wave and the signal attenuation effect. Since the propagation path of the radar wave is relatively long, the echo signal often attenuates as the propagation distance increases. Therefore, the following attenuation model is introduced to represent the variation of the echo signal: Wherein: A(r) is the intensity of the echo signal at a distance of r, A 0 is the initial echo intensity, α is the attenuation factor related to the electromagnetic characteristics of the snow, γ is the attenuation coefficient related to the snow layer thickness, radar wavelength, and frequency.

[0036] Through this model, the intensity variation of the echo signal at different distances can be effectively calculated, and thus the snow depth and humidity can be further deduced.

[0037] As an option, the echo signal can also be further optimized and calibrated by other methods. For example, for the snow characteristics in a specific area, machine learning methods are used to model and train the echo signal to improve the accuracy of echo signal processing. Through this process, more accurate electromagnetic characteristics of the snow cover layer can be obtained, providing more reliable basic data for subsequent steps.

[0038] In a possible implementation, in order to improve the accuracy of signal processing, it is also necessary to further optimize the analysis of the echo signal by combining on-site meteorological data (such as temperature, humidity, wind speed, etc.). These environmental parameters will affect the propagation of the radar wave and the electromagnetic characteristics of the snow, so they need to be considered during the analysis process.

[0039] By analyzing the echo signal in detail, multiple characteristic information about snow cover can be obtained, such as the thickness, humidity, density, etc. of the snow layer. This information is crucial for subsequent snow volume calculation and snow layer feature extraction. Specifically, these features will be used to judge the distribution of snow cover, and further provide a basis for the next step of adaptive sampling and data compression.

[0040] The ultimate goal of this step is to generate an accurate dataset of snow echo signals as the basis for subsequent snow layer analysis, calculation, and prediction. In some embodiments, these data can also be further combined with ground measurement data or meteorological data to improve the accuracy and adaptability of snow volume measurement.

[0041] In this embodiment, by generating and analyzing the echo signal, the snow cover characteristics of the airport pavement can be accurately estimated. Through a variety of signal processing methods (including geometric modeling, reflection coefficient analysis, attenuation models, etc.), reliable snow echo signal data are obtained, which provides basic support for subsequent adaptive sampling, data compression, and deep learning optimization. This process is one of the core steps of the present invention, ensuring the accuracy and efficiency of subsequent technical processes.

[0042] S2. Adaptive Sampling and Data Compression In the foregoing steps, the generation and analysis of the echo signal provide basic data for subsequent snow volume measurement. These data contain various information of the snow layer, such as thickness, humidity, and density, etc. Therefore, in order to ensure the calculation efficiency and accuracy of subsequent steps, it is necessary to further process the echo signal data. In this context, the core task of this step is to dynamically adjust the sampling density of the echo signal through an adaptive sampling method. Through this step, redundant calculations in snow-free areas can be effectively reduced, thereby improving the processing efficiency of the entire measurement system.

[0043] In this embodiment, the intensity of the echo signal is closely related to the density of the snow-covered area. The adaptive sampling method determines the sampling area and sampling density based on the intensity change of the echo signal. Generally, the echo signal in the snow-covered area is stronger, so the sampling density is larger. While in the snow-free area, the echo signal is weaker, so the sampling density is lower. This method avoids redundant calculations in invalid areas and makes the calculation process more efficient.

[0044] Specifically, this embodiment realizes adaptive sampling in the following way. First, by analyzing the intensity of the echo signal, the sampling density D(x,y) of each sampling point is set as: Where: A(x,y) is the intensity of the echo signal at the position (x,y), representing the echo reflection characteristics of this position; γ is a weighting coefficient that controls the relationship between the sampling density and the echo signal intensity.

[0045] From this formula, it can be seen that the sampling density is higher in the regions with stronger echo signals. On the other hand, the sampling density is reduced in the regions with weaker echo signals, thus reducing the computational amount in snow-free areas. This method can ensure fine sampling in snow-covered areas while saving computational resources.

[0046] As an option, the weighting coefficient γ can be adjusted according to specific geographical environments or climatic conditions. For example, in some regions with thinner snow cover or weaker signals, it is necessary to adjust γ to slightly increase the sampling density to improve the accuracy. By adjusting this parameter, it can better adapt to different snow cover environments, thereby further optimizing the sampling process.

[0047] This embodiment also considers the influence of the noise level of the echo signal on the sampling density. In practical applications, radar echo signals are often interfered by noise, which will affect the sampling accuracy and effect. To overcome this problem, the present invention adopts a method of dynamically adjusting γ, such that in the regions with stronger signal noise, the sampling density automatically increases. This adjustment can ensure the stability and accuracy of the sampling process and avoid errors caused by noise.

[0048] Specifically, when the intensity A(x, y) of the echo signal is less than a preset threshold, the system will adjust γ according to the noise level to slightly increase the sampling density to make up for the insufficient sampling caused by weak signals. This dynamic adjustment strategy can ensure a high computational accuracy even in the case of unstable signals.

[0049] Adaptive sampling not only helps to reduce redundant calculations, but also facilitates subsequent data compression and processing. In this embodiment, further steps include compressing the sampled data to reduce the storage and computational burdens. Specifically, the sampled signal is processed by wavelet transform, such that the low-frequency information is retained while the high-frequency noise is effectively removed.

[0050] Wavelet transform is a multi-resolution analysis tool that can decompose a signal into multiple different frequency bands. By performing wavelet transform on the echo signal, the main features of the snow layer can be effectively extracted while removing irrelevant noise. For different snow layer characteristics and echo signal characteristics, a suitable mother wavelet can be selected for processing to improve the signal processing effect and accuracy.

[0051] In practical applications, selecting the Daubechies wavelet as the mother wavelet can provide good time-frequency localization characteristics and is suitable for processing echo signals. The formula for wavelet transform is: Where: $W(a,b)$ is the wavelet transform coefficient at scale $a$ and displacement $b$; $x(t)$ is the input signal, representing the echo signal; $\psi(t)$ is the selected mother wavelet.

[0052] Through this method, the signal is decomposed into multiple components, useful information can be separated, and noise can be removed, thereby greatly reducing the amount of data while retaining the important information of the data.

[0053] This embodiment also includes further processing the compressed data to improve the efficiency of subsequent analysis. Specifically, the data can be corrected by combining meteorological data (such as temperature, humidity, snowfall, etc.), making the final snow accumulation measurement result more accurate. Meteorological data can provide environmental information about the snow layer, further enhancing the accuracy of echo signal processing.

[0054] As a possible implementation, after data compression, the system can calibrate the data in real time and use the calibrated data for subsequent snow accumulation calculation. In this process, the processing combined with meteorological data can further optimize the snow quantification process and ensure the accuracy of the final measurement result.

[0055] This embodiment effectively reduces the calculation of snow-free areas through the adaptive sampling method, improving the calculation efficiency. At the same time, signal compression and noise removal are combined with wavelet transform, which not only reduces the calculation burden but also improves the accuracy of signal analysis. In addition, dynamically adjusting the weighting coefficient $\gamma$ and combining meteorological data further optimize the adaptive sampling process, ensuring high-precision measurement results even under complex meteorological conditions. These technical means complement each other, ensuring the superior performance of the present invention in airport runway snow accumulation measurement.

[0056] S3. Wavelet Transform and Feature Extraction In the foregoing steps, the snow echo signal was optimized through adaptive sampling, and signal compression and preprocessing were completed, laying a foundation for subsequent feature extraction and data analysis. On this basis, this step further performs signal feature extraction, mainly using wavelet transform technology to perform multi-resolution analysis on the echo signal and extract effective snow accumulation features. This process helps to eliminate noise from the signal and retain the main information of the snow layer, thereby enhancing the accuracy of subsequent analysis.

[0057] Specifically, the snow echo signal contains feature information at different levels. Some signals represent the basic structure of the snow layer, while other signals contain high-frequency noise or other interferences. To effectively separate this information, wavelet transform is adopted in this embodiment. Its advantage lies in being able to process signals at multiple scales and capture the subtle changes in the signals. Wavelet transform can provide localized analysis of time and frequency, and is particularly suitable for analyzing signals with multi-scale features, such as the snow echo signal.

[0058] In this embodiment, during the signal processing, the echo signal is first decomposed by wavelet transform, and the features of multiple frequency bands are extracted. By this method, high-frequency noise can be effectively removed, and the main information of the snow layer is retained in the low-frequency part. Specifically, Daubechies wavelet is selected as the mother wavelet, which has good time-frequency localization characteristics and can effectively capture the changes in the snow echo signal.

[0059] The mathematical principle of wavelet transform is as follows: Where: W(a,b) is the wavelet transform coefficient at scale a and displacement b; x(t) is the input signal, representing the echo signal; ψ(t) is the selected mother wavelet for signal decomposition.

[0060] Generally, Daubechies wavelet is selected because it has good resolution when processing signals, especially when dealing with signals with local features. This makes Daubechies wavelet very suitable for extracting local features in the snow echo signal, especially for capturing relatively subtle changes in the signal (such as small changes in the snow layer).

[0061] As an option, other types of wavelets can also be selected for signal decomposition in this embodiment. Specifically, which mother wavelet to choose can be adjusted according to the characteristics of different echo signals. For example, if the characteristics of the snow echo signal change strongly in time, Morlet wavelet will be selected, which has strong localization characteristics in both frequency and time.

[0062] During the data processing, after the signal undergoes wavelet transform, information at multiple scales can be extracted. This information can not only effectively remove noise, but also help further identify the characteristics of the snow, such as the thickness, humidity of the snow layer, and the transition between different snow layers.

[0063] Specifically, the low-frequency part obtained after wavelet transform represents the main features of snow cover, while the high-frequency part mainly contains noise and detailed information. To ensure the accuracy and efficiency of the analysis, the present invention adopts a strategy of removing high-frequency noise, only retaining the low-frequency signal, and further optimizing these features through subsequent deep learning processing. In this way, through efficient feature extraction, the complexity of subsequent calculations can be reduced, and the accuracy of the final snow volume calculation can be improved.

[0064] In some embodiments, the denoising process also needs to be combined with noise estimation techniques to further optimize the results of wavelet transform. For example, median filtering or adaptive filtering methods are used to remove burst noise in the echo signal, thereby further enhancing the stability and accuracy of feature extraction.

[0065] In another possible implementation, the wavelet transform results can also be combined with other signal processing techniques, such as Fourier transform or wavelet packet decomposition, in order to process signals in different frequency bands more meticulously. Through these means, the extraction accuracy of snow cover layer features can be improved, ensuring the reliability of measurement results.

[0066] Furthermore, after wavelet transform, the system automatically removes the high-frequency part, and the remaining low-frequency part is used for subsequent processing. In this process, the extracted low-frequency signal can be used to analyze the overall change trend of the snow cover layer, while the high-frequency signal mainly reflects the tiny fluctuations or noise of the snow layer. Therefore, retaining the low-frequency signal is crucial for the estimation of snow volume.

[0067] This embodiment performs multi-resolution analysis on the echo signal through wavelet transform, which can effectively remove high-frequency noise and retain the main feature information of snow cover. By selecting an appropriate mother wavelet (such as Daubechies wavelet), precise decomposition of the signal can be achieved, thereby providing effective features for subsequent deep learning optimization and snow volume calculation. This process not only improves the data processing efficiency but also enhances the accuracy of subsequent analysis, providing reliable data support for snow measurement on airport runways.

[0068] S4. Deep learning optimization In the foregoing steps, multi-resolution analysis is performed on the snow echo signal through wavelet transform, and the main features of the signal are effectively extracted. These processed signals provide an important data basis for subsequent snow layer feature extraction. To further improve the accuracy of feature extraction and achieve accurate identification of snow-covered areas, this step adopts a convolutional neural network (CNN) for optimization. Through an automated learning process, CNN can extract representative snow cover features from complex echo signals, which are then used for snow volume estimation and area division.

[0069] In this embodiment, the CNN is mainly used to further analyze the signal after wavelet transform processing. Specifically, the echo signal after wavelet transform contains key features of the snow-covered area, such as snow layer thickness, humidity and other information. However, the expression of these information is relatively complex, so it needs to be further processed and classified by a deep learning model. Through training, the CNN can identify the complex snow layer features in the echo signal and distinguish them from non-snow layer areas, thus providing an accurate area division for the final snow accumulation calculation.

[0070] Specifically, the CNN structure consists of multiple convolutional layers and pooling layers. The convolutional layer is used to extract local features from the input signal, while the pooling layer is used to reduce the dimension of these features, thereby improving the calculation efficiency and preventing overfitting. The convolution operation performs a sliding window process on the input signal through a filter (or convolution kernel), and extracts features by gradually learning the hierarchical structure in the signal.

[0071] In this embodiment, the size of the convolution kernel in the convolutional layer is 3×3, and this size has shown good performance in most snow echo signal analyses. The convolution operation can extract local features of the signal, and these features represent the local changes of the signal, which can effectively capture the subtle changes in the snow-covered area. In addition, the pooling layer uses a 2×2 max pooling operation. The role of the pooling layer is to extract the maximum value from each local area, thereby reducing the data volume, reducing the calculation complexity, and at the same time retaining important features.

[0072] In this embodiment, the CNN extracts multi-level features of the signal layer by layer through the combination of multiple convolutional layers and pooling layers. Specifically, the first convolutional layer mainly extracts basic features in the signal, such as edges, textures, etc.; subsequent convolutional layers gradually extract higher-level features, such as the specific shape of the snow layer or other snow-related features. The convolution kernel of each layer is continuously adjusted during the training process to learn the important information in the signal to the greatest extent.

[0073] When training the CNN, a large number of labeled training data are used, and these data include echo signals of snow cover under different weather conditions. Through learning these data, the CNN can gradually identify the snow cover features in the signal and finally automatically classify the echo signal into snow-covered areas and non-snow-covered areas. The loss function L during the training process can be expressed as: where y i is the actual label, indicating whether it is a snow-covered area at position i; is the predicted label output by the CNN network; N is the total number of training samples.

[0074] By optimizing the loss function L, the network can gradually adjust the weights through the gradient descent method, thereby improving the recognition accuracy of the model for snow-covered areas.

[0075] As an option, during the training process, parameters such as the number of layers of the CNN, the size of the convolutional kernels, and the pooling method can be adjusted according to the different characteristics of the dataset, thereby improving the accuracy and adaptability of the model. In some special environments, such as areas where the snow cover is very thin or the snowfall is uneven, the CNN can enhance its processing ability by adjusting the size of the convolutional kernels or introducing more convolutional layers.

[0076] In a possible implementation, to further improve the robustness of the model, data augmentation techniques can also be introduced. The data augmentation techniques simulate various different environmental conditions and snow layer states by performing operations such as randomly rotating, scaling, and cropping the training data, thereby improving the adaptability of the CNN to different snow layer states.

[0077] In addition, the output of the CNN is a binary classification result, that is, classifying the input signal into snow-covered areas and non-snow-covered areas. Based on such classification results, the system can further estimate the snow accumulation amount based on the predicted snow-covered areas, calculate the snow depth in different areas, and finally obtain the total snow accumulation amount on the airport runway.

[0078] In this embodiment, as a deep learning method, the CNN can effectively extract the characteristic information of snow from the echo signal processed by wavelet transform. Through multi-level convolution and pooling, the CNN gradually extracts the complex features in the signal, realizing the automatic recognition of snow-covered areas. Combining the loss function and the gradient descent optimization algorithm, the CNN can continuously optimize its recognition ability, thereby providing an accurate area division for subsequent snow accumulation amount calculation. In this way, the present invention can accurately identify and measure the snow layer under different meteorological conditions, improving the efficiency and accuracy of the overall system.

[0079] S5. Fast Fourier Transform (FFT) Acceleration In the foregoing steps, the characteristics in the snow echo signal were effectively extracted through the convolutional neural network (CNN), and the classification of the signal and the recognition of the snow layer area were successfully completed. To further improve the calculation efficiency and accuracy in the snow detection process, this step uses the Fast Fourier Transform (FFT) technology to accelerate the processing of the echo signal. The FFT can transform the signal from the time domain to the frequency domain, making the signal processing more efficient, especially in filtering high-frequency noise in the signal and extracting signal characteristics, which has significant advantages.

[0080] In this embodiment, FFT is used to accelerate the spectral analysis of echo signals. By converting the time-domain signal into a frequency-domain signal, the system can quickly identify different frequency components in the echo signal, and then effectively extract the characteristics of the snow cover layer. During the signal processing, especially when the echo signal contains a lot of high-frequency noise, FFT can decompose and filter the signal in the frequency domain, remove the irrelevant noise, and enhance the frequency components related to the snow cover, thereby improving the accuracy and speed of signal processing.

[0081] Specifically, for the frequency-domain analysis of the snow echo signal, first perform a Fourier transform on the echo signal to convert its representation from the time domain to the frequency domain. The advantage of the frequency domain is that the characteristics of many signals can be more clearly shown in the frequency domain, especially for the characteristics of the snow cover layer with relatively stable frequencies. Through FFT, the key frequencies in the signal can be quickly identified, the characteristics of the snow can be recognized, and the high-frequency noise can be filtered out.

[0082] In this embodiment, the calculation formula of FFT is: where X(k) is the frequency-domain signal; x(n) is the time-domain signal; N is the length of the signal; k is the frequency index.

[0083] Through FFT, each frequency component in the echo signal can be converted through the corresponding frequency index kkk. The signal in the frequency domain can not only more clearly display the characteristics of the snow cover layer, but also help to identify the noise components irrelevant to the snow in the signal. Generally, the echo signal of the snow cover layer will generate relatively stable characteristics in certain frequency ranges, while the high-frequency noise is generated in other frequency bands, and it can be effectively separated by FFT.

[0084] As an option, the present invention can also further process the frequency-domain signal in combination with a band-pass filter to ensure that only the frequency components related to the snow cover are retained, and further remove the noise components irrelevant to the snow cover. This process optimizes the frequency-domain signal by adjusting the frequency range of the filter, thereby improving the accuracy of subsequent analysis.

[0085] In another possible implementation, in order to further enhance the stability and reliability of the signal, the system can also adjust the parameters of FFT according to the specific characteristics of the echo signal. For example, for some special characteristics of the snow cover layer, more refined analysis will be carried out in a specific frequency range to ensure that the characteristics of the snow cover can be accurately captured in a complex environment. These adjustments can be dynamically adjusted based on real-time environmental data, so that FFT is more adaptable to snow volume measurement under different meteorological conditions.

[0086] Specifically, after the FFT processing, the frequency-domain signal will be transmitted to the subsequent analysis module for further processing. The key in this processing lies in how to identify the effective snow accumulation characteristics and convert them into measurement results such as the thickness or humidity of the snow layer. In some embodiments, the frequency-domain features extracted by FFT will be input into other machine learning models for further pattern recognition and prediction to improve the accuracy of snowfall measurement.

[0087] In this embodiment, by applying the Fast Fourier Transform (FFT) to perform frequency-domain analysis on the echo signal, the efficiency of signal processing can be significantly improved. FFT can not only help identify the characteristics of the snow accumulation layer in the frequency domain but also effectively remove high-frequency noise, enhancing the stability and accuracy of the signal. By combining technologies such as band-pass filters, the signal processing process is further optimized, making the subsequent snow accumulation measurement more efficient and accurate. This technical solution provides strong technical support for snow accumulation monitoring under complex meteorological conditions, ensuring the real-time and accuracy of snow accumulation measurement on airport runways.

[0088] S6. Parallel Computing and Real-Time Feedback In the foregoing steps, the echo signal has been efficiently processed in the frequency domain by the Fast Fourier Transform (FFT), removing noise and enhancing the characteristics of the snow accumulation layer. This step provides reliable signal support for the subsequent snowfall measurement and makes the entire processing process more efficient. However, with the increasing real-time requirement for snow accumulation detection, a single computing resource cannot meet the demand for processing a large amount of data. Therefore, the key task of this step is to use parallel computing technology to distribute the processing tasks to multiple computing nodes to ensure real-time and accuracy while processing large-scale data.

[0089] In this embodiment, a parallel computing and distributed computing framework, such as MapReduce, is adopted to process a large amount of echo signal data. Through parallel computing, the system can split the signal data into multiple subtasks and distribute these subtasks to multiple computing units for processing, thus greatly improving the computing efficiency.

[0090] Specifically, in the foregoing steps, the FFT technology has performed frequency-domain processing on the echo signal, and the amount of processed data is often very large. To accelerate the computing process, the system will split the processed data into multiple parts, each part being processed by an independent computing node. Through this distributed computing, the system can complete the processing of a large amount of data in a relatively short time, thereby improving the real-time of snowfall measurement.

[0091] Generally, the amount of data in the echo signal is large, and it is often difficult for a single processing unit to quickly complete the calculation task. Through parallel computing, the signal processing tasks can be distributed to multiple computing units, and each computing unit is only responsible for a part of the data, greatly shortening the calculation time. Specifically, when using a distributed computing framework, the data will be transferred between different nodes, and each node executes the FFT operation and other related calculation tasks in parallel. These nodes can be multi-core processors, cluster computing nodes, or virtual machines in a cloud computing environment. The calculation tasks processed by each node communicate through the network, and finally the processing results are merged to obtain the complete snow volume detection result.

[0092] As an option, the MapReduce computing framework is adopted in this embodiment, which effectively realizes the parallel processing of data by dividing the tasks into a Map stage and a Reduce stage. In the Map stage, the data is split into multiple segments, and each segment is processed by an independent computing node. Each computing node performs FFT processing, feature extraction, and other analysis tasks on the data. Then, in the Reduce stage, these results will be summarized and merged to generate the final snow layer detection result.

[0093] In a possible implementation, the GPU acceleration technology can also be combined, and the computing tasks are handed over to the graphics processing unit (GPU) for processing. GPUs usually have high parallel processing capabilities and can process a large amount of data in a short time. By executing computing tasks such as FFT on the GPU, the computing efficiency of the system can be further improved.

[0094] This embodiment also involves optimizing the cooperation between computing nodes. To ensure data consistency and processing speed during parallel computing, the system dynamically schedules tasks according to the load conditions of the nodes. For example, when the computing capacity of a certain node reaches saturation, the task will be reassigned to an idle computing node, thus ensuring the efficient use of computing resources and the smooth progress of the computing process.

[0095] Specifically, during the processing, the system can monitor the computing progress of each computing node in real time to ensure the balanced distribution of tasks. If a certain node encounters a processing bottleneck, the system will automatically adjust the task allocation strategy to avoid the real-time nature of the entire processing process being affected by the computing bottleneck.

[0096] In another implementation, the system can also ensure that the entire computing task can continue to execute even if a certain computing node fails through a fault tolerance mechanism. During the distributed computing process, if a certain node fails, the system will take over the task through a backup node to ensure that the computing process does not interrupt. Such a mechanism enhances the reliability and stability of the system.

[0097] In this embodiment, all parallel computing tasks will ultimately generate a merged result, which contains key information such as the area of the snow-covered region, the thickness of the snow layer, and the humidity. To ensure that the system can output measurement results in real time, the final snow accumulation data will be transmitted to the control center or relevant systems through a real-time feedback mechanism. In this way, the airport management can timely understand the snow accumulation situation on the runway and then make corresponding decisions.

[0098] As an option, the real-time feedback mechanism can display the measurement results through a visual interface, enabling managers to intuitively view the snow accumulation situation in each area and formulate the priority of snow removal operations based on this data. In addition, through the cloud computing platform, the system can achieve remote monitoring of data and multi-region collaboration, further enhancing the intelligent level of snow accumulation monitoring.

[0099] Through the parallel computing and distributed computing framework, this embodiment can efficiently process a large amount of echo signal data, ensuring that the system maintains real-time performance while operating efficiently. By adopting means such as the MapReduce framework and GPU acceleration technology, the computing efficiency is greatly improved, ensuring that the snow accumulation measurement process can be completed in a short time and accurate results can be provided. The real-time feedback mechanism in this process enables the airport to respond quickly when the snow situation changes, ensuring the safety of flights and the smooth operation of the airport. Snow Volume Calculation and Real-Time Feedback In the foregoing steps, the processing efficiency of echo signals has been effectively improved through parallel computing and distributed computing frameworks, and the data has been accurately analyzed and processed. As the processing process progresses step by step, the obtained snow layer data is ready for snow volume calculation. In this step, the system will calculate the snow volume based on the results obtained in the foregoing steps and transmit the results to relevant systems through a real-time feedback mechanism, providing accurate snow volume data for airport managers to support the decision-making of snow removal operations and flight scheduling.

[0100] In this embodiment, when calculating the snow volume, the system calculates the snow accumulation volume on the airport runway by comprehensively analyzing key features such as the thickness and humidity of the snow layer. The calculation of the snow accumulation volume not only depends on the intensity of the echo signal but is also closely related to the aforementioned weather data (such as temperature, humidity, and snowfall). By combining these factors, the actual thickness of the snow accumulation layer can be predicted more accurately, and the specific snow accumulation volume can be calculated.

[0101] Specifically, the calculation formula for the snow accumulation volume can be expressed as: Where: S is the snow accumulation volume, with the unit of cubic meters; A i is the area of the i-th region, with the unit of square meters; hi is the snow depth of the i-th area, in meters; ρ i is the snow density of the i-th area, in kilograms per cubic meter; N is the total number of areas, usually referring to the multiple areas into which the airport pavement is divided.

[0102] In some embodiments, the thickness h of the snow i and the density ρ i are obtained from the feature extraction and analysis results of the echo signal in the foregoing steps. The density of the snow layer is usually an environmental variable, which is closely related to the humidity, temperature, and snowfall of the snow. By comprehensively considering these factors, the snow volume of each area can be accurately calculated and summarized to obtain the snow volume of the entire airport pavement.

[0103] As an option, in order to further improve the calculation accuracy, the system can also introduce data fusion technologies such as Kalman filtering to perform real-time fusion on data from different sources (such as radar signal data and meteorological data). This process can effectively eliminate the noise between different data sources and further optimize the calculation result of the snow volume.

[0104] Specifically, when calculating the snow volume, first calculate the snow depth of each area to obtain the height of the snow layer in each area. These height values come from the echo signal features extracted by wavelet transform and CNN network in the previous steps, and are further compared with the actual measured data to ensure the accuracy of the calculation result. At the same time, the density ρ of the snow i can be estimated through a meteorological model or prior data, especially in an environment where the temperature and humidity change greatly, the change of density is more significant.

[0105] In a possible implementation, in order to improve the accuracy of snow calculation under extreme meteorological conditions, the system can perform adaptive adjustment according to historical meteorological data. For example, when the snowfall is large, the system will increase the density calculation of wet snow to improve the calculation accuracy.

[0106] In this embodiment, the snow volume calculation result is quickly fed back to the system on the basis of real-time processing, so that the airport management can immediately grasp the snow condition of the airport pavement. To ensure real-time performance and response speed, the system forwards the snow volume calculation result to the snow removal operation system or flight scheduling system through a real-time feedback mechanism to provide decision support.

[0107] Specifically, the real-time feedback system transmits snow accumulation data to the monitoring platform or control center, and managers can schedule snow removal operations and flight arrangements based on this data. In addition, the system can also generate visual charts of snow accumulation data, providing more intuitive information to help decision-makers quickly evaluate the runway conditions. Through the visual interface, managers can clearly see the snow accumulation situation in each area, and then optimize the priority and resource allocation of snow removal operations.

[0108] As an option, to enhance the operability and flexibility of the system, the system can provide multiple feedback methods, such as mobile device notifications, emails, or system alarms. Especially during large-scale snow accumulation, the system can notify relevant personnel in real time to avoid flight delays and airport operation problems caused by snow condition changes.

[0109] In this embodiment, by combining snow accumulation calculation with a real-time feedback mechanism, it is ensured that while the system efficiently processes data, it can transmit accurate snow accumulation information in real time. By introducing different meteorological data, echo signal characteristics, and calculation formulas, the calculation results of snow accumulation can be more accurate. The application of the real-time feedback mechanism ensures that managers can understand the snow accumulation situation in the first time and then make timely responses. This process can effectively ensure flight safety and improve airport operation efficiency.

[0110] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for measuring snow on airport pavement based on GB-ArcSAR, characterized in that: The following steps are involved: S1. Generate the echo signal of the airport pavement using GB-ArcSAR; S2. Analyze the echo signal, establish a snow echo signal model, and deduce the electromagnetic reflection characteristics of snow; S3. Based on the reflection intensity of the echo signal, the sampling density is adjusted by using an adaptive sampling method, focusing on sampling the snow-covered area and reducing the calculation of the snow-free area; S4. Use wavelet transform to perform multi-resolution analysis on the echo signal, extract the main features of the signal and remove noise, thereby reducing data redundancy; S5. Use convolutional neural network to extract and classify snow echo signals to identify snow layer areas; S6. Use fast Fourier transform to accelerate echo signal processing and improve computational efficiency through frequency domain conversion; S7. Combine the echo signal data with the weather data to calculate the amount of snow on the pavement and provide real-time feedback.

2. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The adaptive sampling method comprises: According to the strength of the echo signal, set the sampling density of each sampling point Where A(x,y) is the echo signal strength at the position (x,y), and γ is the weighting coefficient, which controls the relationship between the sampling density and the echo signal strength; The value of the weighting coefficient γ is dynamically adjusted according to the variation range and noise level of the echo signal.

3. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The steps of wavelet transform include: Perform wavelet transform on the echo signal x(t) to obtain multi-scale wavelet transform coefficients Wherein W(a,b) is the wavelet transform coefficient under scale a and displacement b, ψ(t) is the mother wavelet, x(t) is the echo signal, and the mother wavelet is the Daubechies wavelet, which has the optimal time-frequency localization characteristic and is suitable for processing echo signals.

4. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The convolutional neural network comprises the following steps: Use training sample data to train the network and learn the echo signal characteristics of snow and non-snow areas; The convolution layer and pooling layer are used to extract the snow layer features in the echo signal and output the snow layer classification results; The convolutional neural network includes two convolutional layers and two pooling layers, wherein the convolution kernel size of the convolutional layer is 3×3, and the pooling layer adopts a 2×2 maximum pooling operation.

5. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The loss function L of the convolutional neural network is: Among them, y i is the actual label, is the convolutional neural network prediction label, and N is the number of training samples.

6. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The fast Fourier transform comprises: Perform frequency domain conversion on the echo signal and use the Fourier transform algorithm to convert the time domain signal into the frequency domain signal, thereby speeding up the processing of the echo signal and reducing the computational complexity; The fast Fourier transform calculation formula is: Among them, X(k) is the frequency domain signal, x(n) is the time domain signal, N is the signal length, and k is the frequency index.

7. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 6, characterized in that: The fast Fourier transform is used to accelerate the spectrum analysis of the echo signal, especially to remove high-frequency noise and enhance the snow layer characteristics in the signal.

8. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The method further comprises the following steps: Using parallel computing methods, the snow echo signal data is divided into multiple subtasks and assigned to multi-core processors or distributed computing environments for parallel processing, further improving the processing speed; The parallel computing method includes using a MapReduce computing framework to perform distributed processing on echo signal data to accelerate the snow volume calculation process.

9. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 8, characterized in that: The parallel computing method adopts multi-thread processing to ensure real-time snow detection when processing large amounts of data.

10. The method for measuring snow on an airport pavement based on GB-ArcSAR according to claim 1, characterized in that: The echo signal data is combined with weather data, and the amount of snow on the airport pavement is calculated through a model. The weather data includes temperature, humidity, and snowfall information. By combining with the echo signal data, the amount of snow is accurately calculated.

Images