Highway agglomerate fog prediction method and system based on multi-source remote sensing fusion, and storage medium

Abstract

The invention discloses a highway agglomerate fog prediction method and system based on multi-source remote sensing fusion and a storage medium, and relates to the technical field of agglomerate fog prediction. The method comprises the following steps: acquiring meteorological satellite remote sensing image data and video monitoring image data of a target area; carrying out fusion processing on the meteorological satellite remote sensing image data and the video monitoring image data based on a fusion model of a generative adversarial network to obtain a fused image; establishing a road agglomerate fog data set according to the fused image, and dividing a training set and a test set according to a set proportion; building an agglomerate fog recognition model based on a convolutional neural network and a support vector machine, and training the agglomerate fog recognition model by using the training set to obtain a trained agglomerate fog recognition model; and carrying out agglomerate fog prediction by using the trained agglomerate fog identification model, and carrying out different-degree cost grading early warning prompt according to a prediction result. According to the method, the road agglomerate fog occurrence probability at the future moment can be efficiently and accurately predicted, and graded early warning in advance is realized.

Description

Technical Field

[0001] This invention relates to the field of fog prediction technology, and in particular to a method, system and storage medium for predicting fog on highways based on multi-source remote sensing fusion. Background Technology

[0002] "Damp fog" is a severe weather phenomenon characterized by sudden onset and extremely low visibility, commonly found on highways. With the continuous construction and development of my country's highway network, vehicle safety has become increasingly prominent. In-depth research into damp fog detection methods is crucial for improving the safety and operational efficiency of the national intelligent transportation system. Traditional damp fog detection methods often rely on wireless sensors and lasers to establish monitoring stations, which suffer from drawbacks such as cumbersome technology, implementation difficulties, and poor economic efficiency. Furthermore, due to the mobile nature of damp fog, the accuracy of damp fog forecasts is even lower in highway scenarios. Therefore, it is urgent for those skilled in the art to propose a high-precision method for detecting and warning of road damp fog. Summary of the Invention

[0003] The purpose of this invention is to provide a method, system, and storage medium for predicting road fog based on multi-source remote sensing fusion, in order to solve the problems mentioned in the background art. This invention can efficiently and accurately predict the probability of road fog occurrence at future times and achieve early graded warnings.

[0004] To achieve the above objectives, the present invention provides the following solution: On one hand, a method for predicting highway fog based on multi-source remote sensing fusion, the specific steps of which include the following:

[0005] Acquire meteorological satellite remote sensing image data and video surveillance image data of the target area;

[0006] The meteorological satellite remote sensing image data and the video surveillance image data are fused using a generative adversarial network fusion model to obtain a fused image.

[0007] A highway fog dataset is created based on the fused images, and the training and test sets are divided according to a set ratio.

[0008] A fog recognition model is constructed based on a convolutional neural network and a support vector machine. The fog recognition model is trained using the training set to obtain the trained fog recognition model.

[0009] The trained fog recognition model is used to predict fog patterns, and different levels of warnings are issued based on the prediction results.

[0010] Preferably, the method further includes preprocessing the meteorological satellite remote sensing image data and the video surveillance image data respectively; performing spatiotemporal registration, outlier removal, and noise reduction on the meteorological satellite remote sensing image data; and performing data cleaning, data enhancement, and video frame extraction on the video surveillance image data.

[0011] Preferably, the fusion model includes an adaptive resolution convolutional Transformer generator and a convolutional Transformer discriminator, wherein the adaptive resolution convolutional Transformer generator and the convolutional Transformer discriminator are a deep supervised generative adversarial network constructed from the convolutional Transformer module.

[0012] Preferably, the video surveillance image data is full-angle video image data acquired by an intelligent dome camera. While acquiring the signal, the video data is compressed, and the video image is subjected to fast minimum sparse projection compression and blind sparse reconstruction through Fourier orthogonal basis, and feature extraction is performed on the reconstructed image.

[0013] Preferably, the fog recognition model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer; wherein, the input layer is used to receive the fused images in the training set, the convolutional layer and the pooling layer are used to extract image features, the fully connected layer is used to map the image features into a one-dimensional data format, and the fully connected layer is connected to a support vector machine classifier. After calculation by the support vector machine classifier, the output layer outputs the fog discrimination result.

[0014] Preferably, the method further includes inferring the area where the fog occurs based on the fog prediction results, and determining traffic warning roads and traffic control measures based on the fog occurrence area.

[0015] Preferably, the real-time status update information of the fog patch is obtained, and the fog patch occurrence area is updated based on the latest information through the fog patch recognition model to generate the real-time adjusted fog patch occurrence area and fog patch generation information.

[0016] On the other hand, a highway fog prediction system based on multi-source remote sensing fusion is provided, including a data acquisition module, a data fusion module, a dataset construction module, a training module, and a fog prediction module; wherein,

[0017] The data acquisition module is used to acquire meteorological satellite remote sensing image data and video surveillance image data of the target area;

[0018] The data fusion module is used to fuse the meteorological satellite remote sensing image data and the video surveillance image data based on a generative adversarial network fusion model to obtain a fused image.

[0019] The dataset construction module is used to build a highway fog dataset based on the fused images and divide it into training and test sets according to a set ratio.

[0020] The training module is used to construct a fog recognition model based on a convolutional neural network and a support vector machine, and to train the fog recognition model using the training set to obtain the trained fog recognition model.

[0021] The fog prediction module is used to predict fog patterns using a trained fog recognition model and to provide different levels of warning alerts based on the prediction results.

[0022] Finally, a computer-readable storage medium is provided, on which a computer program is stored, characterized in that, when the computer program is run on a computer, the computer causes the computer to execute the aforementioned method for predicting highway fog based on multi-source remote sensing fusion.

[0023] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects: Based on generative adversarial networks, potential relationships between different source data are constructed through the fusion of multi-source data, which improves the fusion capability and generalization capability. In addition, the fusion of convolutional neural network and support vector machine models for fog detection and classification can effectively improve the overall performance of the model, and has high classification accuracy and robustness in complex foggy road environments. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart of the method of the present invention;

[0026] Figure 2 This is a diagram illustrating the fog formation process of the present invention;

[0027] Figure 3 This is a system structure diagram of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The purpose of this invention is to provide a method for predicting highway fog based on multi-source remote sensing fusion, such as... Figure 1 As shown, the specific steps include the following:

[0030] S1. Acquire meteorological satellite remote sensing image data and video surveillance image data of the target area;

[0031] S2. A fusion model based on generative adversarial networks is used to fuse meteorological satellite remote sensing image data and video surveillance image data to obtain a fused image.

[0032] S3. Establish a highway fog dataset based on the fused images, and divide it into training and test sets according to a set ratio;

[0033] S4. Construct a fog recognition model based on convolutional neural network and support vector machine, and train the fog recognition model using the training set to obtain the trained fog recognition model;

[0034] S5. Use the trained fog recognition model to predict fog patterns and provide different levels of warning prompts based on the prediction results.

[0035] Dense fog is a localized fog phenomenon characterized by its sudden onset, localized nature, and short lifespan. It typically occurs in areas with complex terrain, high humidity, and large temperature variations, especially near highway bridges, low-lying areas, or rivers. When it occurs on highways, it can easily cause traffic accidents. The formation process of dense fog is as follows... Figure 2 As shown, the formation of fog patches is the result of the combined effects of atmospheric physics and meteorological conditions. Its core mechanism lies in the condensation of water vapor in the air into tiny water droplets under low-temperature conditions, reducing visibility to below 200 meters. The formation of fog patches is influenced by a combination of meteorological and environmental factors. Specifically, these factors can be divided into two main categories: meteorological factors and environmental factors. Meteorological factors mainly involve the supply of water vapor in the air, temperature changes, and air pressure. Environmental factors affect the accumulation and diffusion of fog droplets. Accurately understanding the interactions between these factors during the construction of prediction models helps improve the accuracy of fog patch predictions.

[0036] Furthermore, meteorological satellite remote sensing image data and video surveillance image data of the target area are acquired. Fog features are extracted through the fusion of meteorological and environmental factors to further predict fog patterns. The video surveillance image data consists of full-angle video images acquired by an intelligent dome camera. To address the large storage requirements of the video images and to better establish a data sample set, the video data is compressed simultaneously with signal acquisition. Fast minimum sparse projection compression and blind sparse reconstruction of the video images are performed using Fourier orthogonal bases, and feature extraction is performed on the reconstructed images.

[0037] After acquiring the data, preprocessing operations are required for both meteorological satellite remote sensing image data and video surveillance image data. For the meteorological satellite remote sensing image data, spatiotemporal registration, outlier removal, and noise reduction are performed. Specifically, spatiotemporal registration involves matching the raster coordinates of the satellite image with the geographic coordinates of the video surveillance image, using highway mileage markers and latitude / longitude as a reference, with the error controlled within 50m. Outlier removal involves removing cloud / noise pixels from the satellite image. For the video surveillance image data, data cleaning, data augmentation, and video frame extraction are performed. Specifically, data cleaning involves checking data consistency and handling invalid and missing values. Data augmentation involves rotating, flipping, and adjusting the brightness of the video images to address the scarcity of fog samples.

[0038] Furthermore, after data preprocessing, a fusion model based on generative adversarial networks is used to fuse meteorological satellite remote sensing image data and video surveillance image data to obtain fused images. The fusion model includes an adaptive resolution convolutional Transformer generator and a convolutional Transformer discriminator, which are deep supervised generative adversarial networks constructed from convolutional Transformer modules.

[0039] Furthermore, the fog recognition model includes an input layer, convolutional layers, pooling layers, fully connected layers, and an output layer. Convolutional layers are responsible for extracting local features from the image. By using multiple convolutional kernels, the network can learn different features, such as edges and textures. These features form the basis for subsequent classification tasks. Pooling layers, typically following convolutional layers, reduce the spatial dimensionality of features, thereby reducing computation. Pooling also makes feature detection more robust and enhances the model's generalization ability. Fully connected layers, located deep within the network, flatten the outputs of convolutional layers and pass them through fully connected layers to generate high-dimensional feature vectors for classification. The fully connected layers map the learned features to the sample label space and seamlessly connect the one-dimensional data output by the fully connected layers to a support vector machine (SVM) classifier for further analysis. The output layer, closely connected to the last fully connected layer, typically uses a normalized exponential function or logistic function to output classification labels, obtaining the fog discrimination result.

[0040] During the training of the fog recognition model, the model was optimized by adjusting the network depth, the number and size of convolutional kernels, and the pooling layer strategy to achieve the best recognition effect; and the performance of the fog recognition model was evaluated using a test set.

[0041] Furthermore, it also includes inferring the area where the fog will occur based on the fog prediction results, and determining traffic warning roads and traffic control measures based on the area where the fog will occur.

[0042] Obtain real-time status updates of the fog patches, update the fog patch occurrence area based on the latest information using the fog patch recognition model, and generate the real-time adjusted fog patch occurrence area and fog patch generation information.

[0043] On the other hand, a highway fog prediction system based on multi-source remote sensing fusion is provided, such as... Figure 3 As shown, it includes a data acquisition module, a data fusion module, a dataset construction module, a training module, and a fog prediction module; among which,

[0044] The data acquisition module is used to acquire meteorological satellite remote sensing image data and video surveillance image data of the target area;

[0045] The data fusion module is used to fuse meteorological satellite remote sensing image data and video surveillance image data based on a generative adversarial network fusion model to obtain a fused image.

[0046] The dataset construction module is used to create a highway fog dataset based on the fused images and divide it into training and test sets according to a set ratio.

[0047] The training module is used to build a fog recognition model based on convolutional neural networks and support vector machines. The fog recognition model is trained using the training set to obtain the trained fog recognition model.

[0048] The fog prediction module is used to predict fog patterns using a trained fog recognition model and to provide different levels of warning alerts based on the prediction results.

[0049] Finally, a computer-readable storage medium is provided on which a computer program is stored, which, when run on a computer, causes the computer to execute a road fog prediction method based on multi-source remote sensing fusion.

[0050] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for predicting highway fog based on multi-source remote sensing fusion, characterized in that, The specific steps include the following: Acquire meteorological satellite remote sensing image data and video surveillance image data of the target area; The meteorological satellite remote sensing image data and the video surveillance image data are fused using a generative adversarial network fusion model to obtain a fused image. A highway fog dataset is created based on the fused images, and the training and test sets are divided according to a set ratio. A fog recognition model is constructed based on a convolutional neural network and a support vector machine. The fog recognition model is trained using the training set to obtain the trained fog recognition model. The trained fog recognition model is used to predict fog patterns, and different levels of warnings are issued based on the prediction results.

2. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 1, characterized in that, It also includes preprocessing the meteorological satellite remote sensing image data and the video surveillance image data respectively; and performing spatiotemporal registration, outlier removal and noise reduction on the meteorological satellite remote sensing image data. The video surveillance image data is cleaned, enhanced, and frame extracted.

3. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 1, characterized in that, The fusion model includes an adaptive resolution convolutional Transformer generator and a convolutional Transformer discriminator, which are deep supervised generative adversarial networks constructed from convolutional Transformer modules.

4. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 1, characterized in that, The video surveillance image data is full-angle video image data acquired by an intelligent dome camera. While acquiring the signal, the video data is compressed. The video image is then subjected to fast minimum sparse projection compression and blind sparse reconstruction using Fourier orthogonal basis, and features are extracted from the reconstructed image.

5. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 1, characterized in that, The fog recognition model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The input layer receives fused images from the training set. The convolutional and pooling layers extract image features. The fully connected layer maps the image features into a one-dimensional data format and is connected to a support vector machine classifier. After calculation by the support vector machine classifier, the output layer outputs the fog discrimination result.

6. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 1, characterized in that, It also includes inferring the area where the fog will occur based on the fog prediction results, and determining traffic warning roads and traffic control measures based on the fog occurrence area.

7. The method for predicting highway fog based on multi-source remote sensing fusion according to claim 6, characterized in that, Obtain real-time status update information of the fog patch, update the fog patch occurrence area based on the latest information through the fog patch recognition model, and generate the real-time adjusted fog patch occurrence area and fog patch generation information.

8. A highway fog prediction system based on multi-source remote sensing fusion, characterized in that, It includes a data acquisition module, a data fusion module, a dataset construction module, a training module, and a fog prediction module; among which, The data acquisition module is used to acquire meteorological satellite remote sensing image data and video surveillance image data of the target area; The data fusion module is used to fuse the meteorological satellite remote sensing image data and the video surveillance image data based on a generative adversarial network fusion model to obtain a fused image. The dataset construction module is used to build a highway fog dataset based on the fused images and divide it into training and test sets according to a set ratio. The training module is used to construct a fog recognition model based on a convolutional neural network and a support vector machine, and to train the fog recognition model using the training set to obtain the trained fog recognition model. The fog prediction module is used to predict fog patterns using a trained fog recognition model and to provide different levels of warning alerts based on the prediction results.

9. A computer-readable storage medium, characterized in that, It stores a computer program, characterized in that when the computer program is run on a computer, the computer causes the computer to execute a highway fog prediction method based on multi-source remote sensing fusion as described in any one of claims 1 to 7.

Images