Hazardous waste pollution accident early warning system and method

Abstract

The invention relates to the field of pollution accident early warning, and particularly discloses a hazardous waste pollution accident early warning system and method, and the method comprises the steps: firstly obtaining an environment remote sensing image collected by a satellite camera, an expressway real-time monitoring video, and expressway accident environment sensing data, and then carrying out the deep learning technology, and finally, through a classifier, whether hazardous waste pollution accident early warning is sent out or not is judged, an alarm is given out in time, and the accuracy and timeliness of accident early warning are improved.

Description

Technical Field

[0001] The present application relates to the field of pollution accident early warning, and more specifically, to a hazardous waste pollution accident early warning system and method. Background Art

[0002] On highways, a large number of vehicles transporting hazardous chemicals pose potential safety risks. Once a leakage accident occurs in these vehicles, it will directly pose a serious threat to the environment and public health, or even be fatal.

[0003] At present, early warning methods for hazardous chemical leaks on highways usually rely on video surveillance or telephone alarm systems. However, these methods often need to wait until the situation develops to a certain stage before issuing an alarm, which may cause environmental pollution or health risks to have already occurred. This passive response mode may delay the implementation of emergency response measures, thereby affecting the control effect of the incident and the effectiveness of emergency treatment.

[0004] Therefore, a hazardous waste pollution accident early warning system and method are desired. Summary of the invention

[0005] In order to solve the above technical problems, this application is proposed. The embodiment of this application provides a hazardous waste pollution accident early warning system and method, which first obtains environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos and highway accident environmental perception data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three, and finally uses a classifier to determine whether to issue a hazardous waste pollution accident early warning, so as to issue an alarm in time and improve the accuracy and timeliness of accident early warning.

[0006] According to one aspect of the present application, a hazardous waste pollution accident early warning system is provided, which includes:

[0007] Hazardous waste pollution data acquisition module, used to obtain environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos and highway accident environment perception data;

[0008] A hazardous waste pollution data extraction module, used to extract hazardous waste pollution associated feature vectors and highway accident environment perception feature vectors from the environmental remote sensing images collected by the satellite camera, the highway real-time monitoring video and the highway accident environment perception data;

[0009] The pollution accident warning classification module is used to determine whether to issue a hazardous waste pollution accident warning based on the hazardous waste pollution association feature vector and the highway accident environment perception feature vector.

[0010] According to another aspect of the present application, a method for early warning of a hazardous waste pollution accident is provided, which comprises:

[0011] Acquire environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos, and highway accident environment perception data;

[0012] Extracting hazardous waste pollution-related feature vectors and highway accident environment perception feature vectors from the environmental remote sensing images collected by the satellite camera, the highway real-time monitoring video, and the highway accident environment perception data;

[0013] Based on the hazardous waste pollution associated feature vector and the highway accident environment perception feature vector, it is determined whether to issue a hazardous waste pollution accident warning.

[0014] Compared with the prior art, the present application provides a hazardous waste pollution accident early warning system and method, which first obtains environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos and highway accident environment perception data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a classifier is used to determine whether to issue a hazardous waste pollution accident early warning, and an alarm is issued in a timely manner to improve the accuracy and timeliness of accident early warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a block diagram of a hazardous waste pollution accident early warning system according to an embodiment of the present application.

[0017] Figure 2 It is a block diagram of a hazardous waste pollution data extraction module in a hazardous waste pollution accident early warning system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram of the environmental perception data feature extraction unit in the hazardous waste pollution accident warning system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram of a pollution accident warning classification module in a hazardous waste pollution accident warning system according to an embodiment of the present application.

[0020] Figure 5 Flow chart of a hazardous waste pollution accident early warning method according to an embodiment of the present application.

[0021] Figure 6 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] Figure 1 Schematic diagram of the hazardous waste pollution accident early warning system of the present application embodiment. Figure 1 As shown, according to the hazardous waste pollution accident early warning system 100 of the embodiment of the present application, it includes: a hazardous waste pollution data acquisition module 110, which is used to obtain environmental remote sensing images collected by satellite cameras, real-time monitoring videos of highways and environmental perception data of highway accidents; a hazardous waste pollution data extraction module 120, which is used to extract hazardous waste pollution-related feature vectors and highway accident environmental perception feature vectors from the environmental remote sensing images collected by satellite cameras, the real-time monitoring videos of highways and the environmental perception data of highway accidents; a pollution accident early warning classification module 130, which is used to determine whether to issue a hazardous waste pollution accident early warning based on the hazardous waste pollution-related feature vectors and the highway accident environmental perception feature vectors.

[0024] In the above-mentioned hazardous waste pollution accident early warning system 100, the hazardous waste pollution data acquisition module 110 is used to obtain environmental remote sensing images, real-time highway monitoring videos, and highway accident environmental perception data collected by satellite cameras. It should be understood that on highways, vehicles that frequently transport hazardous chemicals have potential safety hazards. Once a leakage accident occurs, it may directly pose a major threat to the environment and public health or even a fatal danger. At present, the early warning method for hazardous chemical leakage on highways mainly relies on video monitoring or telephone alarm systems. However, these methods usually need to wait for the situation to develop to a certain extent before issuing an alarm, which may cause environmental pollution or health risks to have occurred. This passive response mode may delay the implementation of emergency response measures, thereby affecting the accident control effect and the effectiveness of emergency treatment. Therefore, in the technical solution of the present application, by obtaining environmental remote sensing images, real-time highway monitoring videos, and highway accident environmental perception data collected by satellite cameras, and combining deep learning technology, the possibility of hazardous waste pollution events is evaluated in real time, and an alarm is issued in time to improve the accuracy and timeliness of the early warning.

[0025] Specifically, satellite cameras provide wide-area environmental remote sensing data that can capture environmental changes and potential hazardous waste emissions over a large area. These images can not only be used for target detection and feature extraction, but also can be used to extract environmental remote sensing features through professional algorithms to analyze environmental changes and anomalies. Real-time highway monitoring videos provide real-time road traffic conditions and accident monitoring data, which can instantly capture various types of accidents occurring on the highway, including traffic accidents and possible hazardous waste leaks. Through the analysis and processing of monitoring videos, the system can extract key monitoring image frames for accident monitoring tracking and feature extraction, as well as further early warning analysis. At the same time, highway accident environmental perception data provides detailed environmental information about the accident site, such as meteorological conditions, traffic flow, surrounding environment and other data. These data are collected in real time through sensing devices and sensors, which can help the system more accurately assess the scope and consequences of the accident and provide important input parameters for the early warning system. The acquisition of these data can not only improve the efficiency and accuracy of early warning of hazardous waste pollution accidents, but also help relevant departments to respond quickly and take effective emergency measures to minimize the impact of pollution accidents on the environment and human health.

[0026] In the above-mentioned hazardous waste pollution accident early warning system 100, the hazardous waste pollution data extraction module 120 is used to extract hazardous waste pollution-related feature vectors and highway accident environmental perception feature vectors from the environmental remote sensing images collected by the satellite camera, the real-time monitoring video of the highway, and the environmental perception data of the highway accident. It should be understood that the extraction of hazardous waste pollution-related feature vectors and highway accident environmental perception feature vectors from the environmental remote sensing images collected by the satellite camera, the real-time monitoring video of the highway, and the environmental perception data of the highway accident is to build an effective early warning system that can detect and respond to hazardous waste pollution events and related highway accidents in a timely manner. Extracting these feature vectors not only enhances the early warning system's monitoring capabilities for hazardous waste pollution and highway accidents, but also provides important support for emergency response and resource scheduling, ensuring that the dual goals of public safety and environmental protection are achieved.

[0027] Figure 2 FIG. 1 is a block diagram of a hazardous waste pollution data extraction module in a hazardous waste pollution accident early warning system according to an embodiment of the present application. Figure 2As shown, in a specific embodiment of the present application, the hazardous waste pollution data extraction module 120 includes: an environmental remote sensing image feature extraction unit 121, used to perform feature extraction on the environmental remote sensing image collected by the satellite camera to obtain an environmental remote sensing feature vector; a highway monitoring feature extraction unit 122, used to perform feature extraction on the real-time monitoring video of the highway to obtain a highway dangerous accident monitoring and tracking feature vector; a hazardous waste pollution feature association unit 123, used to associate the environmental remote sensing feature vector with the highway dangerous accident monitoring and tracking feature vector to obtain the hazardous waste pollution association feature vector; an environmental perception data feature extraction unit 124, used to perform feature extraction on the highway accident environmental perception data to obtain the highway accident environmental perception feature vector.

[0028] It should be understood that satellite cameras provide wide coverage and high-resolution environmental data, which can capture a variety of physical features of the surface, such as topography, land cover type, water distribution, etc. These images usually contain rich visual information, but it is unrealistic to directly use the original image data for analysis and processing because of the huge amount of data and high complexity. Through feature extraction technology, these complex image data can be converted into more informative and actionable feature vectors. The feature extraction process usually includes a variety of algorithms and methods, such as color feature extraction, texture analysis, shape description, etc. These features can abstractly describe important information in the image, such as the classification of surface cover types, the density of vegetation cover, changes in land use, etc.

[0029] Furthermore, feature extraction of real-time monitoring video of highways is performed to obtain feature vectors for monitoring and tracking dangerous accidents on highways in order to achieve effective monitoring and management of traffic safety. As an important traffic artery, highways have large traffic volume, high speed and high accident risk. Real-time monitoring video can provide instant visual information, including the driving status of vehicles, traffic conditions on roads and potential accident locations. However, it is unrealistic to directly use raw video data for real-time analysis and decision-making because the video data is huge and the amount of information is huge. Through feature extraction technology, complex video data can be converted into efficient feature vectors. These feature vectors can capture key information in the video, such as vehicle speed, density, driving direction, lane deviation, emergencies (such as braking, collision, etc.). These feature vectors can not only simplify the complexity of data processing, but also improve the system's ability to detect traffic anomalies and potential accidents. Feature vectors can realize rapid identification and response to the moment of accident, so as to issue alarms in time and initiate emergency response measures to reduce the severity and scope of the accident.

[0030] Furthermore, the purpose of associating the environmental remote sensing feature vectors with the highway dangerous accident monitoring and tracking feature vectors is to establish a comprehensive environmental monitoring and early warning system that can effectively respond to hazardous waste pollution incidents and their potential impacts. The environmental remote sensing feature vectors provide a description of environmental information such as surface cover, vegetation conditions, and land use through image data collected by satellite cameras. These feature vectors can help identify potential environmental anomalies such as hazardous waste stacking and spillage, thereby early detection and response to potential pollution incidents. The highway dangerous accident monitoring and tracking feature vectors provide a description of traffic flow, vehicle status, and possible accident situations through real-time monitoring videos. These feature vectors can monitor and track accidents on highways in real time, including traffic accidents that may lead to hazardous waste leakage, such as collisions or leaks of transport vehicles. Associating the environmental remote sensing feature vectors with the highway dangerous accident monitoring and tracking feature vectors can not only improve the comprehensive capabilities of the environmental monitoring and early warning system, but also effectively prevent and reduce the negative impact of hazardous waste pollution incidents on the environment and society, and promote sustainable development and safe traffic management.

[0031] In particular, the highway accident environment perception data covers a variety of environmental parameters and real-time monitoring information, such as traffic flow, vehicle speed, weather conditions, road conditions, etc. These data are collected in real time through various sensors, which can provide key environmental conditions and accident influencing factors, and provide basic data support for rapid response after the accident. Through feature extraction technology, representative and practical highway accident environment perception feature vectors can be extracted from a large amount of original perception data. These feature vectors include but are not limited to: accurate positioning of the accident location, real-time evaluation of the accident impact area, trend analysis of traffic flow changes, etc. Through the establishment and update of these feature vectors, the system can quickly respond to traffic accidents, reduce traffic congestion after the accident, the incidence of further accidents, and rescue time.

[0032] In a specific embodiment of the present application, the environmental remote sensing image feature extraction unit 121 includes: passing the environmental remote sensing image collected by the satellite camera through an environmental remote sensing target detection network to obtain an environmental remote sensing area of ​​interest; passing the environmental remote sensing area of ​​interest through an environmental remote sensing feature encoder including a significant target detector to obtain the environmental remote sensing feature vector.

[0033] It should be understood that satellite cameras are able to acquire large-scale, high-resolution surface image data, but these data usually contain a large amount of background information and non-target areas, and direct processing and analysis are inefficient. Through an environmental remote sensing target detection network, such as a convolutional neural network based on deep learning, areas of interest in the image, such as buildings, forests, farmland, lakes, etc., can be automatically identified and marked. The target detection network uses advanced image processing technology and trained models to accurately locate and identify various objects and landscape features in complex environments. These networks learn rich image feature expressions during the training process and can effectively distinguish between targets and backgrounds, thereby improving the extraction accuracy and efficiency of the region of interest. Through the target detection network, the time and computing resources for processing non-target areas can be reduced, and data processing efficiency and cost can be optimized. Specifically, the environmental remote sensing target detection network is used to process the environmental remote sensing image collected by the satellite camera using the following target detection formula to obtain the environmental remote sensing region of interest;

[0034] Among them, the target detection formula is:

[0035] ROI=(cls(ψ det ,B),Regr(ψ det ,B))

[0036] Among them, ψ det is the environmental remote sensing image collected by the satellite camera, B is the anchor frame, ROI is the environmental remote sensing region of interest, cls(ψ det ,B) indicates classification, Regr(ψ det ,B) indicates regression.

[0037] Further, the environmental remote sensing regions of interest are processed by an environmental remote sensing feature encoder including a salient object detector, aiming to transform complex image information into compact and representative feature vectors, thereby supporting higher-level environmental monitoring, classification and application. Environmental remote sensing regions of interest are important geographical areas or specific features such as buildings, vegetation cover, water bodies, etc., which are identified and extracted by object detectors. These areas not only reflect the actual conditions of the surface, but may also contain important information related to environmental changes. The application of salient object detectors in environmental remote sensing helps to highlight and extract key regions of interest by identifying and quantifying the most significant feature areas in the image, such as color, texture, shape, etc. These areas may need to be feature encoded before further processing so that they can be expressed as numerical feature vectors. The feature vector captures the multi-dimensional features of each region of interest, such as color histogram, texture features, spatial distribution, etc., making the feature information of each region richer and more specific. Processing environmental remote sensing regions of interest by an environmental remote sensing feature encoder including a salient object detector not only helps to improve the efficiency and accuracy of data processing, but also provides more in-depth and comprehensive data support for a wide range of environmental applications, promoting the development of environmental science and technology. Specifically, each layer of the environmental remote sensing feature encoder including the salient target detector performs the following operations on the input data in the forward pass of the layer: convolution processing is performed on the input data using a first convolution kernel to obtain a first convolution feature map; convolution processing is performed on the first convolution feature map using a second convolution kernel to obtain a second convolution feature map, wherein the size of the first convolution kernel is larger than the size of the second convolution kernel; pooling processing is performed on the second convolution feature map to obtain a pooled feature map; activation processing is performed on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the first convolution neural network model is the environmental remote sensing feature vector, and the input of the first layer of the first convolution neural network model is the environmental remote sensing area of ​​interest.

[0038] In a specific embodiment of the present application, the highway monitoring feature extraction unit 122 includes: extracting the real-time monitoring video of the highway at a predetermined sampling frequency to obtain a plurality of highway dangerous accident monitoring image key frames; passing the plurality of highway dangerous accident monitoring image key frames through a highway dangerous accident monitoring key frame convolutional neural network as a generator to obtain a plurality of highway dangerous accident monitoring generated image key frames; passing the plurality of highway dangerous accident monitoring generated image key frames through a highway dangerous accident monitoring generated key frame feature encoder using a temporal attention mechanism to obtain the highway dangerous accident monitoring tracking feature vector.

[0039] It should be understood that real-time monitoring videos of highways usually contain a large amount of transport vehicles and traffic flow information, as well as possible accident scenes. In order to save storage space and processing costs, as well as to improve data processing efficiency, key frame extraction with a predetermined sampling frequency is an effective strategy. The key frame extraction process selects and saves frames with representative and key information from a continuous video stream by setting a fixed time interval or based on a specific video analysis algorithm. These frames usually include key moments before, during and after the accident, such as abnormal vehicle driving, collision or other emergency situations. Through the extraction of key frames of multiple highway dangerous accident monitoring images, the key frames can effectively identify and capture the moment of the accident, providing a real-time data foundation for the accident monitoring system. These images can help monitoring personnel respond quickly and reduce the time and manpower costs of accident handling.

[0040] Furthermore, multiple highway dangerous accident monitoring image key frames are processed by highway dangerous accident monitoring key frame convolutional neural network as a generator, aiming to generate highway accident monitoring image key frames with higher resolution, richer information and more realism through machine learning and deep learning technology, so as to support more accurate accident analysis, simulation and emergency response. Highway dangerous accident monitoring image key frame convolutional neural network as a generator, based on advanced generative adversarial network (GAN) or variational autoencoder (VAE) architecture, can learn and simulate complex scenes and details when accidents occur, such as vehicle motion trajectory, collision dynamics, etc. This network can generate high-fidelity images similar to but not necessarily identical to actual monitoring images under given input conditions, such as the time series of key frames or specific accident type labels, by training the learned image generation model. These generated image key frames are not just simple copies, but potential features and structures that can be learned from the data, thereby enhancing the information content and expression ability of the image.

[0041] Furthermore, multiple highway dangerous accident monitoring generated image key frames are processed by the highway dangerous accident monitoring generated key frame feature encoder using the temporal attention mechanism, aiming to extract the feature vectors of temporal correlation and important events from the generated image sequence to support the spatiotemporal analysis, pattern recognition and prediction of accident monitoring. The temporal attention mechanism allows the key frame feature encoder to perform weighted processing on each time step in the generated image sequence, identifying and strengthening those frames that are particularly important in the occurrence and development of the accident. This mechanism can effectively distinguish the importance of different time points, thereby improving the expressive power of the feature vector and the accuracy of capturing temporal information. Through the highway dangerous accident monitoring generated key frame feature encoder, the relationship and evolution process between adjacent frames in time in the image sequence can be identified and encoded. This temporally correlated feature vector not only captures the spatial information in the static image, but also includes the dynamic process and temporal trajectory of the accident development. Specifically, adjacent first monitoring key frames and second monitoring key frames are extracted from the multiple highway dangerous accident monitoring image key frames; the first monitoring key frame and the second monitoring key frame are respectively passed through the first convolution layer and the second convolution layer of the highway dangerous accident monitoring key frame feature encoder using the time attention mechanism to obtain a first convolution feature map and a second convolution feature map; the position point multiplication between the first convolution feature map and the second convolution feature map is calculated to obtain a time attention map; the time attention map is input into the Softmax activation function to obtain a time attention feature map; the second monitoring key frame is passed through the third convolution layer of the convolutional neural network model to obtain a third convolution feature map; and, the position point multiplication between the third convolution feature map and the time attention feature map is calculated to obtain the highway dangerous accident monitoring tracking feature vector.

[0042] Figure 3 FIG. 1 is a block diagram of an environmental perception data feature extraction unit in a hazardous waste pollution accident early warning system according to an embodiment of the present application. Figure 3 As shown, in a specific embodiment of the present application, the environmental perception data feature extraction unit 124 includes: an environmental perception data context encoding subunit 1241, which is used to pass the highway accident environmental perception data through a highway accident environmental perception text data context encoder including an embedding layer to obtain a plurality of highway accident environmental perception text feature vectors; an environmental perception text feature encoding subunit 1242, which is used to perform feature encoding on the plurality of highway accident environmental perception text feature vectors to obtain the highway accident environmental perception feature vector.

[0043] It should be understood that highway accident environment perception data covers text information from multiple sources, including traffic condition reports, accident scene descriptions, eyewitness and monitoring system records. These data need to be processed and encoded so that the system can understand and utilize the key information. The highway accident environment perception text data context encoder is able to convert each text sample into a high-dimensional feature vector through an embedding layer and a deep learning model. These feature vectors not only capture the vocabulary and grammatical structure in the text, but also reflect the contextual information of the text, such as the location, time, traffic conditions and related events of the accident. The encoder is able to convert the text description into a semantically rich vector representation that captures the semantic meaning and important features of key information in the accident environment. These vectors can be used for further analysis and processing in automated systems, such as situational awareness, event recognition and accident pattern analysis. Specifically, the highway accident environment perception data is segmented to obtain a word sequence; the embedding layer of the highway accident environment perception text data context encoder including the embedding layer is used to map each word in the word sequence into a word embedding vector to obtain a sequence of word embedding vectors; the converter-based Bert model of the highway accident environment perception text data context encoder including the embedding layer is used to perform global context semantic encoding on the sequence of word embedding vectors to obtain multiple highway accident environment perception text feature vectors.

[0044] Furthermore, feature encoding is performed on multiple highway accident environment perception text feature vectors, aiming to convert the original complex text information into more representative and usable highway accident environment perception feature vectors through effective data processing and feature extraction techniques. Highway accident environment perception text feature vectors may come from a variety of data sources, such as traffic incident reports, eyewitness descriptions, surveillance video annotations, etc. These data need to be processed and analyzed in order to extract key features and important information from the accident environment. The feature encoding process uses machine learning and natural language processing techniques to map each text feature vector to a lower-dimensional, more representative and abstract feature space. This mapping helps to reduce the dimensionality of the data and extract feature representations that can retain the original information to the greatest extent, thereby facilitating subsequent data analysis and model application. The encoding process can effectively compress the original text data, reduce storage and processing costs, while retaining the main content and semantic elements of the text information. These feature vectors are usually easier to manage and analyze than the original text.

[0045] In a specific embodiment of the present application, the environment-aware text feature encoding subunit 1242 includes: two-dimensionally arranging the multiple highway accident environment-aware text feature vectors to obtain a highway accident environment-aware feature matrix; passing the highway accident environment-aware feature matrix through a highway accident environment-aware convolutional neural network as a feature encoder to obtain a highway accident environment-aware feature map; and pooling the highway accident environment-aware feature map to obtain the highway accident environment-aware feature vector.

[0046] It should be understood that the highway accident environment perception text feature vectors usually exist in the form of a single vector, each of which represents a specific text description or data source. These vectors may contain various information, such as the time, location, type, impact and other key data of the accident. By arranging these feature vectors into a two-dimensional matrix in a predefined manner, it is possible to establish a structured relationship and contextual connection between vectors while retaining the independence of each vector. For example, the vectors can be arranged in time series, or arranged by accident type or geographical location to form a larger data structure. The feature matrix integrates the scattered text feature vectors into a unified data structure for easy management and access. This systematic structure helps to effectively organize and manage large amounts of text data and reduce information fragmentation and data redundancy.

[0047] Furthermore, the highway accident environment perception feature matrix is ​​processed by the highway accident environment perception convolutional neural network as a feature encoder, aiming to use the powerful feature extraction ability of the convolutional neural network to further extract high-level, semantically rich feature images from the structured feature matrix to support more accurate and effective accident environment perception and analysis. The highway accident environment perception feature matrix contains key data features extracted from multiple text feature vectors, which cover multiple aspects of the accident, such as time, location, type, impact, etc. These data are organically combined in the form of a matrix, providing rich input information for the neural network. The highway accident environment perception convolutional neural network can effectively extract and abstract the feature matrix in space and time through multi-layer convolution and pooling operations. This processing method can capture the complex relationship between different features in the accident environment, such as the spatial distribution of different accident events, time evolution, and the influence of related factors. Specifically, each layer of the highway accident environment perception convolutional neural network used as a feature encoder performs convolution processing, mean pooling processing based on the local feature matrix and nonlinear activation processing on the input data in the forward transfer of the layer, so that the last layer of the highway accident environment perception convolutional neural network used as a feature encoder outputs the highway accident environment perception feature map, wherein the input of the highway accident environment perception convolutional neural network used as a feature encoder is the highway accident environment perception feature matrix.

[0048] Furthermore, the highway accident environment perception feature map is pooled to reduce the data dimension and extract key information, so as to transform the complex feature image into a compact and representative highway accident environment perception feature vector to support subsequent data analysis, model application and decision support. The highway accident environment perception feature map is a high-level abstract representation extracted from the feature matrix by a convolutional neural network. These feature images reflect important patterns, structures and semantic information in the accident environment, but their dimensions are usually high, which is not conducive to direct application to subsequent machine learning models or decision systems. Pooling operations (such as maximum pooling or average pooling) merge the pixel values ​​of each feature image area into a single representative value by block processing and downsampling the feature image. This operation helps to reduce the complexity of the data while retaining important feature patterns and structural information. Pooling operations can significantly reduce the size and complexity of each feature map, thereby transforming high-dimensional feature maps into lower-dimensional feature vectors. This dimensionality reduction process makes the data easier to process and store, while reducing the complexity of subsequent model calculations.

[0049] In the above-mentioned hazardous waste pollution accident early warning system 100, the pollution accident early warning classification module 130 is used to determine whether to issue a hazardous waste pollution accident early warning based on the hazardous waste pollution associated feature vector and the highway accident environment perception feature vector. It should be understood that the hazardous waste pollution associated feature vector covers key parameters and data in the waste treatment, storage and transportation process, such as waste type, quantity, treatment method, etc. These feature vectors are collected in real time by monitoring systems or sensors, reflecting the potential risks and abnormal conditions in the hazardous waste management process. The highway accident environment perception feature vector includes relevant data of traffic accidents, such as the time, location, type of accident and its impact range. These data are collected through traffic monitoring systems, road sensors or traffic event reports, providing a detailed description of events and environmental changes on highways. Comprehensively analyzing the hazardous waste pollution associated feature vector and the highway accident environment perception feature vector can comprehensively assess the risk of possible hazardous waste pollution events. For example, if high-concentration toxic waste transportation is detected in an area where accidents frequently occur, it may indicate an increase in potential pollution risks.

[0050] Figure 4 FIG. 1 is a block diagram of a pollution accident warning classification module in a hazardous waste pollution accident warning system according to an embodiment of the present application. Figure 4 As shown, in a specific embodiment of the present application, the pollution accident warning classification module 130 includes: a hazardous waste pollution accident warning feature fusion unit 131, which is used to fuse the hazardous waste pollution association feature vector and the highway accident environment perception feature vector to obtain a hazardous waste pollution accident warning judgment classification feature vector; a hazardous waste pollution accident warning feature fusion unit 132, which is used to perform metric compatibility perception based on the network parameter matrix on the hazardous waste pollution accident warning judgment classification feature vector to obtain an optimized hazardous waste pollution accident warning judgment classification feature vector; a hazardous waste pollution accident warning judgment classification unit 133, which is used to pass the optimized hazardous waste pollution accident warning judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a hazardous waste pollution accident warning.

[0051] It should be understood that the fusion of hazardous waste pollution association feature vectors and highway accident environmental perception feature vectors to obtain hazardous waste pollution accident warning judgment classification feature vectors aims to improve the prediction and judgment capabilities of hazardous waste pollution accidents by combining key data features from different fields, thereby effectively reducing potential risks to the environment and public health. Comprehensive analysis of hazardous waste pollution association feature vectors and highway accident environmental perception feature vectors can more comprehensively assess the risk level of possible hazardous waste pollution accidents. For example, combining highway accident data and waste transportation routes can identify possible pollution risks in areas where traffic accidents frequently occur.

[0052] Further, in the technical solution of the present application, considering that environmental remote sensing images, highway monitoring videos and accident environment perception data all contain a large amount of information, these complex data can be converted into more concise and more representative optimized hazardous waste pollution accident early warning judgment classification feature vectors through feature mining. Considering that when the neural network model performs feature mining on the input data based on the network parameter matrix, if the inherent implicit association between the network output features and the network parameter matrix can be considered, the prior information of the neural network model can be further used to modulate the characteristic manifold of the hazardous waste pollution accident early warning judgment classification feature vector to reduce the possibility of special instance disturbances of its characteristic values. Specifically, the prior information of the neural network model is used to modulate the characteristic manifold of the hazardous waste pollution accident early warning judgment classification feature vector in order to further optimize and improve the classification performance. The neural network model can capture the complex structure and correlation information of the data features by learning the patterns and rules of a large amount of data. In this case, by using the prior information of the neural network model to modulate the characteristic manifold of the hazardous waste pollution accident early warning judgment classification feature vector, the generalization ability and robustness of the model in handling the hazardous waste pollution accident early warning judgment task can be effectively enhanced.

[0053] Among them, the metric compatibility perception based on the network parameter matrix is ​​performed on the hazardous waste pollution accident early warning judgment classification feature vector to obtain the optimized hazardous waste pollution accident early warning judgment classification feature vector, including: calculating the mean and variance of the hazardous waste pollution accident early warning judgment classification feature vector; calculating the difference between the eigenvalue of the predetermined position of the hazardous waste pollution accident early warning judgment classification feature vector and the mean of the hazardous waste pollution accident early warning judgment classification feature vector to obtain a first intermediate calculation value; multiplying the network parameter matrix by the hazardous waste pollution accident early warning judgment classification feature vector and then subtracting it from the hazardous waste pollution accident early warning judgment classification feature vector by position to obtain a first intermediate vector, and calculating the Frobenius norm of the first intermediate vector to obtain a first norm value; the first The norm value is added to the variance of the hazardous waste pollution accident early warning judgment classification feature vector and then the square root is taken to obtain a second intermediate calculation value; the first intermediate calculation value is divided by the second intermediate calculation value to obtain a third intermediate calculation value; the squares of each position in the hazardous waste pollution accident early warning judgment classification feature vector are calculated and then accumulated to obtain a first total value; the square root of the first total value is taken and then divided by the scale of the network parameter matrix to calculate the natural exponential function value to obtain a fourth intermediate calculation value, wherein the scale of the network parameter matrix is ​​equal to the product between the width value of the network parameter matrix and the height value thereof; the fourth intermediate calculation value is multiplied by the weighted hyperparameter and then added to the third intermediate calculation value to obtain the eigenvalue of the predetermined position of the optimized hazardous waste pollution accident early warning judgment classification feature vector.

[0054] Among them, the metric compatibility perception of the classification feature vector of the early warning judgment of the hazardous waste pollution accident is performed based on the network parameter matrix to obtain the optimized classification feature vector of the early warning judgment of the hazardous waste pollution accident, including:

[0055]

[0056] Among them, V is the classification feature vector of the dangerous waste pollution accident early warning judgment, M is the network parameter matrix, β represents the weighted hyperparameter, v i is the eigenvalue of the i-th eigenvalue in the classification feature vector for early warning judgment of hazardous waste pollution accidents, N is the scale of the network parameter matrix, the scale of the network parameter matrix is ​​equal to the product of the width value and the height value of the network parameter matrix, L is the length of the classification feature vector for early warning judgment of hazardous waste pollution accidents, ||·|| F The Frobenius norm of the classification feature vector representing the early warning judgment of hazardous waste pollution accidents, represents matrix multiplication, μ represents the mean of the feature set composed of all eigenvalues ​​of the classification feature vector for early warning judgment of hazardous waste pollution accidents, σ 2represents the variance of the feature set composed of all eigenvalues ​​of the classification feature vector of the hazardous waste pollution accident early warning judgment, v' i The i-th eigenvalue in the optimized hazardous waste pollution accident early warning judgment classification feature vector.

[0057] That is, in the technical solution of the present application, considering that when the neural network model performs feature mining on the input data based on the network parameter matrix, if the inherent implicit association between the network output feature and the network parameter matrix can be further considered, the prior information of the neural network model can be further used to modulate the characteristic manifold of the hazardous waste pollution accident early warning judgment classification feature vector to reduce the possibility of the occurrence of special instance disturbances of its characteristic values. Based on this, in a preferred embodiment, the hazardous waste pollution accident early warning judgment classification feature vector is subjected to metric compatibility perception based on the network parameter matrix, which first maps the hazardous waste pollution accident early warning judgment classification feature vector to the high-dimensional feature space where the network parameter matrix is ​​located, and uses the compatibility perception factor between the mapping result and the original input as the inherent implicit association descriptor between the network output feature and the network parameter matrix. Then, the statistical characteristics of the feature group of the hazardous waste pollution accident early warning judgment classification feature vector are used as the inherent modal information of the hazardous waste pollution accident early warning judgment classification feature vector, and the inherent implicit association descriptor is used as the instance disturbance differential adjustment variable to perform data dynamic standardization on the attribute values ​​of each position of the hazardous waste pollution accident early warning judgment classification feature vector. At the same time, the class-standardized modulus data of the hazardous waste pollution accident early warning judgment classification feature vector is used as a bias item to constrain the acceptable range of the population distribution of the hazardous waste pollution accident early warning judgment classification feature vector. In this way, the attribute influence of the interference distribution of abnormal samples can be suppressed as much as possible, thereby improving the expression accuracy of the hazardous waste pollution accident early warning judgment classification feature vector.

[0058] Furthermore, the optimized classification feature vectors for early warning judgment of hazardous waste pollution accidents are obtained by integrating the hazardous waste pollution association feature vectors and the highway accident environment perception feature vectors. These vectors contain key data features in multiple fields such as waste management and traffic accidents, such as waste type, accident time, location, etc., reflecting various situations and background information of potential pollution incidents. The classifier can process and analyze the feature vector data from the monitoring system in real time and quickly generate corresponding classification results. This immediacy can help decision makers and emergency response teams make corresponding early warning decisions in the shortest time.

[0059] In summary, the embodiment of the present application first obtains environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos, and highway accident environment perception data, and then uses deep learning technology to perform feature extraction and correlation analysis on the three. Finally, a classifier is used to determine whether to issue a hazardous waste pollution accident warning, and to issue an alarm in a timely manner to improve the accuracy and timeliness of accident warnings.

[0060] As described above, the hazardous waste pollution accident early warning system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the hazardous waste pollution accident early warning system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the hazardous waste pollution accident early warning system 100 can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the hazardous waste pollution accident early warning system 100 can also be one of the many hardware modules of the terminal device.

[0061] Alternatively, in another example, the hazardous waste pollution accident warning system 100 and the terminal device may also be separate devices, and the hazardous waste pollution accident warning system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0062] Figure 5 Flow chart of the method for early warning of hazardous waste pollution accidents according to an embodiment of the present application. Figure 5 As shown, according to the hazardous waste pollution accident warning method of the embodiment of the present application, it includes: S110, acquiring environmental remote sensing images collected by satellite cameras, real-time monitoring videos of highways and environmental perception data of highway accidents; S120, extracting hazardous waste pollution-related feature vectors and highway accident environmental perception feature vectors from the environmental remote sensing images collected by satellite cameras, the real-time monitoring videos of highways and the environmental perception data of highway accidents; S130, judging whether to issue a hazardous waste pollution accident warning based on the hazardous waste pollution-related feature vectors and the highway accident environmental perception feature vectors.

[0063] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned hazardous waste pollution accident early warning method have been referred to above. Figures 1 to 4 The description of the hazardous waste pollution accident early warning system has been introduced in detail, and therefore, its repeated description will be omitted.

[0064] Below, reference Figure 6 To describe an electronic device according to an embodiment of the present application.

[0065] like Figure 6As shown, the electronic device 10 includes: a processor 11 and a memory 13. The processor 11 and the memory 13 are connected, such as through a bus 12. Optionally, the electronic device 10 may also include a transceiver 14. It should be noted that in actual applications, the transceiver 14 is not limited to one, and the structure of the electronic device 10 does not constitute a limitation on the embodiments of the present invention.

[0066] The processor 11 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of the present invention. The processor 11 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0067] The bus 12 may include a path for transmitting information between the above components. The bus 12 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 12 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 6 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0068] The memory 13 is used to store a computer program corresponding to the hazardous waste pollution accident early warning method of the above embodiment of the present invention, and the computer program is controlled and executed by the processor 11. The processor 11 is used to execute the computer program stored in the memory 13 to implement the content shown in the above method embodiment.

[0069] Among them, the electronic device 10 includes but is not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The electronic device 10 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0070] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0071] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0072] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0073] In the description of the present invention, it is to be understood that the terms “center”, “longitudinal”, “lateral”, “length”, “width”, “thickness”, “up”, “down”, “front”, “back”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inside”, “outside”, “clockwise”, “counterclockwise”, “axial”, “radial”, “circumferential”, etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0074] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0075] In the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements, unless otherwise clearly defined. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0076] In the present invention, unless otherwise clearly specified and limited, a first feature being "above" or "below" a second feature may mean that the first and second features are in direct contact, or the first and second features are in indirect contact through an intermediate medium. Moreover, a first feature being "above", "above" or "above" a second feature may mean that the first feature is directly above or obliquely above the second feature, or simply means that the first feature is higher in level than the second feature. A first feature being "below", "below" or "below" a second feature may mean that the first feature is directly below or obliquely below the second feature, or simply means that the first feature is lower in level than the second feature.

[0077] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A hazardous waste pollution accident early warning system, characterized in that: include: Hazardous waste pollution data acquisition module, used to obtain environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos and highway accident environment perception data; A hazardous waste pollution data extraction module, used to extract hazardous waste pollution associated feature vectors and highway accident environment perception feature vectors from the environmental remote sensing images collected by the satellite camera, the highway real-time monitoring video and the highway accident environment perception data; The pollution accident warning classification module is used to determine whether to issue a hazardous waste pollution accident warning based on the hazardous waste pollution association feature vector and the highway accident environment perception feature vector.

2. The hazardous waste pollution accident early warning system according to claim 1 is characterized in that: The hazardous waste pollution data extraction module includes: An environmental remote sensing image feature extraction unit, used for extracting features from the environmental remote sensing image collected by the satellite camera to obtain an environmental remote sensing feature vector; A highway monitoring feature extraction unit, used for extracting features from the real-time monitoring video of the highway to obtain a highway dangerous accident monitoring and tracking feature vector; A hazardous waste pollution feature association unit, used to associate the environmental remote sensing feature vector with the highway dangerous accident monitoring and tracking feature vector to obtain the hazardous waste pollution association feature vector; The environmental perception data feature extraction unit is used to extract features from the highway accident environmental perception data to obtain the highway accident environmental perception feature vector.

3. The hazardous waste pollution accident early warning system according to claim 2 is characterized in that: The environmental remote sensing image feature extraction unit comprises: The environmental remote sensing image collected by the satellite camera is passed through an environmental remote sensing target detection network to obtain an environmental remote sensing area of ​​interest; The environmental remote sensing region of interest is passed through an environmental remote sensing feature encoder including a salient target detector to obtain the environmental remote sensing feature vector.

4. The hazardous waste pollution accident early warning system according to claim 3 is characterized in that: The highway monitoring feature extraction unit comprises: Extracting the real-time monitoring video of the expressway at a predetermined sampling frequency to obtain a plurality of key frames of the expressway dangerous accident monitoring images; The plurality of highway dangerous accident monitoring image key frames are respectively passed through a highway dangerous accident monitoring key frame convolutional neural network as a generator to obtain a plurality of highway dangerous accident monitoring generated image key frames; The plurality of highway dangerous accident monitoring generated image key frames are generated through a highway dangerous accident monitoring generated key frame feature encoder using a temporal attention mechanism to obtain the highway dangerous accident monitoring tracking feature vector.

5. The hazardous waste pollution accident early warning system according to claim 4 is characterized in that: The environment perception data feature extraction unit includes: An environment perception data context encoding subunit, used for passing the highway accident environment perception data through a highway accident environment perception text data context encoder including an embedding layer to obtain a plurality of highway accident environment perception text feature vectors; The environment-aware text feature encoding subunit is used to perform feature encoding on the multiple highway accident environment-aware text feature vectors to obtain the highway accident environment-aware feature vector.

6. The hazardous waste pollution accident early warning system according to claim 5 is characterized in that: The environment-aware text feature encoding subunit includes: Arranging the plurality of highway accident environment perception text feature vectors in two dimensions to obtain a highway accident environment perception feature matrix; Passing the highway accident environment perception feature matrix through a highway accident environment perception convolutional neural network as a feature encoder to obtain a highway accident environment perception feature map; The highway accident environment perception feature map is pooled to obtain the highway accident environment perception feature vector.

7. The hazardous waste pollution accident early warning system according to claim 6 is characterized in that: The pollution accident early warning classification module includes: A hazardous waste pollution accident early warning feature fusion unit, used to fuse the hazardous waste pollution association feature vector and the highway accident environment perception feature vector to obtain a hazardous waste pollution accident early warning judgment classification feature vector; A hazardous waste pollution accident early warning feature optimization unit is used to perform metric compatibility perception on the hazardous waste pollution accident early warning judgment classification feature vector based on the network parameter matrix to obtain an optimized hazardous waste pollution accident early warning judgment classification feature vector; The hazardous waste pollution accident early warning judgment classification unit is used to pass the optimized hazardous waste pollution accident early warning judgment classification feature vector through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a hazardous waste pollution accident early warning.

8. The hazardous waste pollution accident early warning system according to claim 7 is characterized in that: The hazardous waste pollution accident early warning feature optimization unit includes: Calculating the mean and variance of the hazardous waste pollution accident early warning judgment classification feature vector; Calculating the difference between the characteristic value of the predetermined position of the dangerous waste pollution accident early warning judgment classification characteristic vector and the mean value of the dangerous waste pollution accident early warning judgment classification characteristic vector to obtain a first intermediate calculation value; Multiplying the network parameter matrix by the hazardous waste pollution accident early warning judgment classification feature vector, and then subtracting the matrix from the hazardous waste pollution accident early warning judgment classification feature vector by position to obtain a first intermediate vector, and calculating the Frobenius norm of the first intermediate vector to obtain a first norm value; Add the first norm value to the variance of the hazardous waste pollution accident early warning judgment classification feature vector and then take the square root to obtain a second intermediate calculation value; dividing the first intermediate calculated value by the second intermediate calculated value to obtain a third intermediate calculated value; Calculating the squares of each position in the hazardous waste pollution accident early warning judgment classification feature vector and then accumulating and summing them to obtain a first total value; Calculate a natural exponential function value after taking the square root of the first sum value and dividing it by the scale of the network parameter matrix to obtain a fourth intermediate calculation value, wherein the scale of the network parameter matrix is ​​equal to the product of the width value of the network parameter matrix and the height value thereof; The fourth intermediate calculated value is multiplied by the weighted hyperparameter and then added to the third intermediate calculated value to obtain the characteristic value of the predetermined position of the optimized hazardous waste pollution accident early warning judgment classification characteristic vector.

9. A method for early warning of hazardous waste pollution accidents, characterized in that: include: Acquire environmental remote sensing images collected by satellite cameras, real-time highway monitoring videos, and highway accident environment perception data; Extracting hazardous waste pollution-related feature vectors and highway accident environment perception feature vectors from the environmental remote sensing images collected by the satellite camera, the highway real-time monitoring video, and the highway accident environment perception data; Based on the hazardous waste pollution associated feature vector and the highway accident environment perception feature vector, it is determined whether to issue a hazardous waste pollution accident warning.

10. The method for early warning of hazardous waste pollution accidents according to claim 9, characterized in that: Extracting hazardous waste pollution-related feature vectors and highway accident environment perception feature vectors from the environmental remote sensing images collected by the satellite camera, the highway real-time monitoring video, and the highway accident environment perception data includes: Performing feature extraction on the environmental remote sensing image collected by the satellite camera to obtain an environmental remote sensing feature vector; Extracting features from the real-time monitoring video of the expressway to obtain a monitoring and tracking feature vector for dangerous accidents on the expressway; Associating the environmental remote sensing feature vector with the highway dangerous accident monitoring and tracking feature vector to obtain the hazardous waste pollution associated feature vector; Feature extraction is performed on the highway accident environment perception data to obtain the highway accident environment perception feature vector.

Images