Multi-scene monitoring and early warning method and system based on big data

Abstract

The invention relates to the technical field of expressway traffic safety, in particular to a multi-scene monitoring and early warning method and system based on big data, and the method comprises the steps: collecting monitoring image information and position information obtained by a vehicle-mounted ETC video all-in-one machine on an expressway, carrying out the preprocessing of the monitoring image information, and obtaining a video frame image; judging whether the change degree of the video frame image at the next moment is greater than a set degree value or not based on the video frame image at the previous moment; if yes, obtaining a feature information matrix; analyzing the feature information matrix by adopting an image analysis model to obtain an image analysis result; analyzing the abnormal video frame image by adopting an abnormal analysis model to obtain an abnormal analysis result; and generating early warning information according to the abnormal analysis result and the position information, and sending the early warning information to a driver. By adopting the method, the problems that an expressway monitoring device cannot monitor each area of the expressway in real time and cannot give an early warning in time when an abnormal condition occurs can be solved.

Description

Technical Field

[0001] The present invention relates to the field of highway traffic safety technology, and in particular to a multi-scenario monitoring and early warning method and system based on big data. Background Art

[0002] With the rapid development of society and the expansion of the scale of highways, there are more and more vehicles and people traveling on highways. However, existing highways are equipped with monitoring devices or capture devices at fixed locations, which can only monitor fixed areas, and cannot monitor all areas of the highway in real time, and cannot understand the specific details. In addition, when abnormal situations occur, timely warnings cannot be given. Although people traveling on highways can check the route status through mobile phones or in-vehicle maps, and spontaneously upload accident information when encountering abnormal situations, the accident information uploaded by people is random, and the timeliness of the accident information cannot be guaranteed, and thus it is impossible to send warning information to other people traveling on the highway in a timely manner.

[0003] Therefore, the present invention provides a multi-scenario monitoring and early warning method and system based on big data in combination with the vehicle-mounted ETC video integrated device to solve the above problems. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a multi-scenario monitoring and early warning method and system based on big data, so as to solve the problem that the above-mentioned highway monitoring device cannot monitor various areas of the highway in real time and cannot provide timely warning when abnormal situations occur.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is: In the first aspect, a multi-scenario monitoring and early warning method based on big data includes: Real-time collection of monitoring image information and location information obtained by the ETC video all-in-one device on the highway.

[0006] The monitoring image information is preprocessed to obtain a preprocessed image; and the preprocessed image is converted into a video frame image arranged in chronological order.

[0007] Based on the video frame image at the previous moment, it is determined whether the degree of change of the video frame image at the next moment is greater than the set degree value; if so, a cluster analysis is performed on the video frame image at the next moment to determine the division type of the video frame image at the next moment, and a feature extraction model is used to extract feature information in the video frame image at the next moment to form a feature information matrix.

[0008] According to the division type, the corresponding information analysis unit in the image analysis model is used to analyze the feature information matrix to obtain the image analysis result.

[0009] When there is abnormal information in the image analysis result, an abnormal video frame image of a set time period starting from a later moment is obtained, and the abnormal video frame image is analyzed using an abnormal analysis model to obtain an abnormal analysis result.

[0010] Generate warning information based on abnormal analysis results and location information; and obtain driver information that is on the highway within a set range and has not driven to the location information; and send warning information to the driver through the driver information.

[0011] Preferably, the method of judging whether the degree of change of the video frame image at the next moment is greater than a set degree value based on the video frame image at the previous moment includes: converting the video frame image into a pixel information set of the corresponding channel through the R, G, and B three-color channels, and storing them in chronological order; obtaining the pixel information set at the previous moment and the pixel information set at the next moment; calculating the degree of change between the pixel information set at the next moment and the pixel information set at the previous moment, and the degree of change value The calculation formula is: , in, is the i-th pixel in the pixel information set of the R channel at the next moment, is the i-th pixel in the pixel information set of the R channel at the previous moment, is the i-th pixel in the pixel information set of the G channel at the next moment, is the i-th pixel in the pixel information set of the G channel at the previous moment, is the i-th pixel in the pixel information set of the B channel at the next moment, is the i-th pixel in the pixel information set of the B channel at the previous moment; judge the degree of change Is it greater than the set value?

[0012] Preferably, the clustering analysis of the video frame image at the next moment to determine the division type of the video frame image at the next moment includes: converting the video frame image at the next moment into an initial clustering data set via a preset converter; and analyzing the initial clustering data set using a clustering model constructed based on historical image data to obtain the division type.

[0013] Preferably, the extraction process of the feature extraction model includes: dividing the video frame image at the next moment into multiple irregular block images; according to the division type, screening out the block images that meet the corresponding division type from each block image to obtain multiple key analysis images; using a feature extractor to extract feature information of each key analysis image to form multiple feature information matrices.

[0014] Preferably, the processing process of the image analysis model includes: according to the division type, each feature information matrix is ​​input into the corresponding information analysis unit; each information analysis unit judges whether the feature information matrix meets the set standards of the preset feature information library according to its own preset feature information library, and when it does not meet the set standards, an abnormal block image is obtained, as well as an image analysis result containing abnormal block image information.

[0015] Preferably, the setting standard of the preset feature information library includes a standard setting interval of each element in the feature information matrix and a setting variation interval between the element and surrounding elements.

[0016] Preferably, the analysis process of the abnormal analysis model includes: according to the abnormal block image information, images containing abnormal block images are screened from abnormal video frame images to obtain abnormal images; when the number of abnormal images exceeds a preset number, each abnormal image is analyzed based on a preset abnormal information library corresponding to the division type, and a first membership value of each abnormal image and a second membership value corresponding to the number of abnormal images are determined; the first membership value and the second membership value are processed by a weighted average method, and an average membership value is calculated, and the level corresponding to the average membership value is used as the abnormal level; and an abnormal analysis result is formed according to the division type, the abnormal level and each abnormal image.

[0017] Preferably, the screening process for obtaining abnormal images includes: based on the feature information matrix of the abnormal block image, determining whether the feature information matrix converted from the abnormal video frame image contains a feature similarity matrix whose similarity exceeds a set similarity; if so, marking the abnormal video frame image as an abnormal image.

[0018] Preferably, the information analysis unit includes at least a weather analysis unit, a vehicle analysis unit, a high-speed ground analysis unit, and a surrounding environment analysis unit.

[0019] In the second aspect, the present invention provides a multi-scenario monitoring and early warning system based on big data, the multi-scenario monitoring and early warning system including a data acquisition module, a data preprocessing module, an image analysis module, an anomaly analysis module, a data storage module and an early warning module; the multi-scenario monitoring and early warning system is used to implement a monitoring and early warning method as described in any of the above methods through mutual cooperation between the modules.

[0020] The beneficial effects of the present invention are: 1. The present invention collects monitoring images in front of the vehicle on the highway in real time based on the vehicle-mounted ETC video integrated machine, obtains video frame images arranged in chronological order that can be analyzed after preprocessing the monitoring images; based on the video frame images at the previous moment, judges the degree of change of the video frame images at the next moment, when the degree of change is greater than the set degree value, cluster analysis is performed on the video frame images at the next moment, determines the division type of the video frame at the next moment, and uses a feature extraction model to process the video frame images at the next moment to obtain a feature information matrix; then, according to the division type, the corresponding information analysis unit in the image analysis model is used to analyze the feature information matrix to obtain an image analysis result; detects whether there is abnormal information in the image analysis result, and when abnormal information is detected in the image analysis result, uses the abnormal analysis model to analyze the abnormal video frame images of the set time period obtained from the next moment to obtain an abnormal analysis result; finally, generates warning information according to the abnormal analysis result and the position information, and sends the warning information to the drivers within the set range who are on the highway and have not driven to the position information to remind the drivers. The present invention can solve the problem that the highway monitoring device cannot monitor various areas of the highway in real time and cannot provide timely warning when abnormal situations occur, thereby improving driving safety on the highway.

[0021] 2. The present invention first determines whether the degree of change of the video frame image at the next moment is greater than the set degree value based on the video frame image at the previous moment. If so, the division type of the video frame image at the next moment is determined and the characteristic information matrix corresponding to the video frame image at the next moment is obtained; then the information analysis unit corresponding to the division type in the image analysis model is used to analyze the characteristic information matrix to obtain the image analysis result; when there is abnormal information in the image analysis result, the abnormal video frame image is analyzed using the abnormal analysis model to obtain the abnormal analysis result. The present invention can ensure the accuracy of the image analysis result through the multi-layer analysis method composed of the above model, thereby improving the accuracy of the early warning information. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 The present invention is a schematic flowchart of a multi-scenario monitoring and early warning method based on big data.

[0023] Figure 2 It is a schematic flow chart of the analysis process of the abnormal analysis model of the present invention. DETAILED DESCRIPTION

[0024] The following will refer to the attached Figure 1 and attached Figure 2The embodiments of the present invention are described in detail. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.

[0025] A multi-scenario monitoring and early warning method based on big data, as shown in the attached Figure 1 As shown, the following steps are included: Step S11: Real-time collection of monitoring image information and location information obtained by the ETC video integrated device on the vehicle on the highway.

[0026] Preferably, in step S11, when the vehicle carrying the ETC video integrated machine enters the expressway, the monitoring image information obtained by the ETC video integrated machine is uploaded to the data storage module in the background in real time and subsequently processed. Different ETC video integrated machines obtain different monitoring images, and can obtain monitoring images of various scenes. The location information is the current location information of the vehicle, which can be the location information uploaded by GPS or other vehicle navigation.

[0027] Step S12: pre-process the monitoring image information to obtain a pre-processed image; and convert the pre-processed image into a video frame image arranged in chronological order.

[0028] Step S13: Based on the video frame image at the previous moment, determine whether the change degree of the video frame image at the next moment is greater than the set degree value; if so, execute step S14, if not, execute step S13.

[0029] Specifically, when the degree of change of the video frame image at the next moment is less than or equal to the set degree value, the video frame image at the next moment is used as the new video frame image at the previous moment, and the video frame image at the next moment is used as the video frame image at the next moment, that is, based on the video frame image at the previous moment, it is judged in chronological order whether the degree of change of the video frame image at the next moment is greater than the set degree value.

[0030] Step S14: performing cluster analysis on the video frame image at the next moment, determining the division type of the video frame image at the next moment, and using a feature extraction model to extract feature information in the video frame image at the next moment to form a feature information matrix.

[0031] Preferably, the preset division types include at least weather, highway surface, vehicle, and highway surrounding environment.

[0032] Step S15: According to the division type, the corresponding information analysis unit in the image analysis model is used to analyze the feature information matrix to obtain the image analysis result.

[0033] Specifically, the information analysis unit includes at least a weather analysis unit, a vehicle analysis unit and a high-speed ground analysis unit. Each analysis unit analyzes the characteristic information matrix corresponding to its type to obtain an initial image analysis result, and then integrates the initial image analysis results to obtain a final image analysis result.

[0034] Step S16: when there is abnormal information in the image analysis result, obtain the abnormal video frame image of the set time period starting from the next moment, and use the abnormal analysis model to analyze the abnormal video frame image to obtain the abnormal analysis result.

[0035] Preferably, the set time period is 10s, 20s, 30s or 60s. The specific set time period can be determined according to actual conditions, for example, according to the actual vehicle speed.

[0036] Step S17: Generate warning information based on the abnormal analysis results and the location information; and obtain the driver information on the expressway within the set range and not driving to the location information; send warning information to the driver through the driver information.

[0037] Specifically, in step S17, after generating warning information based on the abnormal analysis results and location information, the system connects with various communication platforms, vehicle suppliers, transportation platforms, etc. to obtain information about drivers within the set range who are on the highway and have not driven to the abnormal analysis results; then, according to the contact information of the driver information, warning information is sent to these drivers to achieve the warning effect. The form of warning information includes but is not limited to voice and text.

[0038] In this embodiment, the present invention collects monitoring images in front of the vehicle on the highway in real time based on the vehicle-mounted ETC video integrated device, and obtains video frame images arranged in chronological order that can be analyzed after preprocessing the monitoring images; based on the video frame images at the previous moment, the degree of change of the video frame images at the next moment is judged. When the degree of change is greater than the set degree value, the video frame images at the next moment are clustered and analyzed to determine the division type of the video frame at the next moment, and the video frame images at the next moment are processed using a feature extraction model to obtain a feature information matrix; then, according to the division type, the feature information matrix is ​​analyzed using the corresponding information analysis unit in the image analysis model to obtain an image analysis result; it is detected whether there is abnormal information in the image analysis result, and when it is detected that there is abnormal information in the image analysis result, the abnormal video frame images of the set time period obtained from the next moment are analyzed using the abnormal analysis model to obtain an abnormal analysis result; finally, an early warning message is generated according to the abnormal analysis result and the position information, and the early warning message is sent to the driver who is within the set range, on the highway and has not driven to the position information, so as to remind the driver. The present invention can solve the problem that the highway monitoring device cannot monitor various areas of the highway in real time and cannot provide timely warning when abnormal situations occur, thereby improving driving safety on the highway.

[0039] And the present invention firstly judges whether the degree of change of the video frame image at the next moment is greater than the set degree value based on the video frame image at the previous moment, and if so, determines the division type of the video frame image at the next moment and obtains the characteristic information matrix corresponding to the video frame image at the next moment; then uses the information analysis unit corresponding to the division type in the image analysis model to analyze the characteristic information matrix to obtain the image analysis result; when there is abnormal information in the image analysis result, uses the abnormal analysis model to analyze the abnormal video frame image to obtain the abnormal analysis result. The present invention can ensure the accuracy of the image analysis result through the multi-layer analysis method composed of the above model, thereby improving the accuracy of the early warning information.

[0040] In the prior art, although people traveling on the highway can check the route status through mobile phones or in-vehicle maps, and when encountering abnormal situations, people can spontaneously upload accident information, but the accident information uploaded by people is random, and the timeliness of the accident information cannot be guaranteed, and thus it is impossible to send early warning information to other people traveling on the highway in time. Compared with the prior art, the present invention can analyze the monitoring images collected by the ETC video all-in-one machine in real time, obtain early warning information in time, and send early warning information to drivers of other vehicles in time, thereby improving the safety of driving on the highway.

[0041] In one embodiment of the present invention, the specific processing process of step S13 includes: converting the video frame image into pixel information sets of corresponding channels through the three color channels of R, G, and B, and storing them in chronological order; obtaining the pixel information set of the previous moment and the pixel information set of the next moment; calculating the change degree value between the pixel information set of the next moment and the pixel information set of the previous moment, and the change degree value The calculation formula is: , in, is the i-th pixel in the pixel information set of the R channel at the next moment, is the i-th pixel in the pixel information set of the R channel at the previous moment, is the i-th pixel in the pixel information set of the G channel at the next moment, is the i-th pixel in the pixel information set of the G channel at the previous moment, is the i-th pixel in the pixel information set of the B channel at the next moment, is the i-th pixel in the pixel information set of the B channel at the previous moment; judge the degree of change Is it greater than the set value?

[0042] Preferably, the setting degree value is between 70 and 100, which can be adjusted according to actual needs.

[0043] In this embodiment, the present invention determines the degree of change of the video frame image at the latter moment compared to the video frame image at the previous moment by comparing the pixel differences in the pixel information set of the R, G, and B three-color channels at the previous moment and the pixel information set at the latter moment, thereby accurately judging the degree of image change and improving the accuracy of judgment.

[0044] In one embodiment of the present invention, step S14 performs cluster analysis on the video frame image at the next moment to determine the division type of the video frame image at the next moment, including: converting the video frame image at the next moment into an initial clustering data set via a preset converter; and analyzing the initial clustering data set using a clustering model constructed based on historical image data to obtain the division type.

[0045] Specifically, the preset converter can be a pixel converter, that is, converting the video frame image at the next moment into a corresponding pixel set or pixel matrix; or the preset converter is a converter built based on a convolutional neural network, etc. The clustering model built based on historical image data can be a clustering model built after processing the historical image data using a k-means clustering algorithm; or a clustering model built using a hierarchical clustering algorithm, and there is no restriction on the algorithm of the constructed clustering model, so that the clustering model can analyze and process the initial clustering data set to determine the division type. When the algorithm is determined, the algorithm is not changed, so that the clustering process is fixed and no new clustering results will appear.

[0046] In this embodiment, there is at least one type of division type for a single video frame image, and the number of division types is not limited.

[0047] In one embodiment of the present invention, the extraction process of the feature extraction model in step S14 includes: dividing the video frame image at the next moment into multiple irregular block images; according to the division type, screening out the block images that meet the corresponding division type from each block image to obtain multiple key analysis images; using a feature extractor to extract feature information of each key analysis image to form multiple feature information matrices.

[0048] In this embodiment, the present invention uses an image segmentation method to segment the video frame image at the next moment into a plurality of irregular block images according to the color and pixel differences in the video frame image at the next moment, so that the block images have only one division type. Then, according to the division type, each block image can be assigned to a corresponding division type list or set to obtain a key analysis image of each division type. Then, a feature extractor is used to extract feature information of each key analysis image to obtain a plurality of feature information matrices.

[0049] Preferably, the image segmentation algorithm can be a convolutional neural network algorithm, a 3D convolutional neural network algorithm, or a convolutional neural network algorithm based on an attention mechanism, so as to segment the video frame image at the next moment into multiple irregular block images. The feature extractor can be a feature extractor constructed by a convolutional neural network or a recurrent neural network.

[0050] The present invention can accurately obtain the feature information matrix corresponding to each division type through the above-mentioned feature extraction model, thereby improving the accuracy of feature extraction.

[0051] In one embodiment of the present invention, the processing process of the image analysis model in step S15 includes: according to the division type, each feature information matrix is ​​input into the corresponding information analysis unit; each information analysis unit judges whether the feature information matrix meets the set standards of the preset feature information library according to its own preset feature information library, and when it does not meet the set standards, an abnormal block image is obtained, and an image analysis result containing abnormal block image information is obtained.

[0052] Furthermore, the setting standard of the preset feature information library includes a standard setting interval of each element in the feature information matrix and a setting variation interval between the element and surrounding elements.

[0053] Preferably, the abnormal block image information includes a link to the abnormal block image, a size of the abnormal block image, a storage location of the abnormal block image, and the like.

[0054] Specifically, a preset feature information library is established based on the historical image data corresponding to each information analysis unit collected on the highway. The preset information library contains standard matrices corresponding to various feature information matrices, as well as the upper and lower fluctuation ranges of each element in the standard matrix, the correlation range between each element, etc., which can accurately determine whether the feature information matrix meets one or more of the standard matrices in the preset information library of each information analysis unit. When it meets one, the feature information matrix meets the set standard of the preset feature information library; when the feature information matrix does not meet all the standard matrices in the preset information library of each information analysis unit, the feature information matrix is ​​an abnormal matrix, and the key analysis image corresponding to the feature information matrix is ​​an abnormal block image, thereby obtaining an image analysis result containing an abnormal block image. In the image analysis result, normal block images are not recorded, and are only represented by words or symbols such as other key analysis images are normal.

[0055] The present invention can accurately determine whether the feature information matrix corresponding to the key analysis image meets the set standards of the preset feature information library corresponding to each unit through the above-mentioned image analysis model, and accurately determine the abnormal key analysis image to obtain the abnormal analysis image; then obtain the image analysis result containing the abnormal block image information, thereby improving the accuracy of the image analysis result.

[0056] In one embodiment of the present invention, as shown in the attached Figure 2 As shown, the analysis process of the abnormal analysis model includes the following steps: Step S21: According to the abnormal block image information, images containing abnormal block images are screened from abnormal video frame images to obtain abnormal images.

[0057] Step S22: When the number of abnormal images exceeds a preset number, each abnormal image is analyzed based on a preset abnormal information library corresponding to the classification type to determine a first membership value of each abnormal image and a second membership value corresponding to the number of abnormal images.

[0058] Step S23: Process the first membership value and the second membership value using a weighted average method to calculate an average membership value, and use the level corresponding to the average membership value as the abnormal level.

[0059] Step S24: generating an abnormality analysis result according to the classification type, abnormality level and each abnormal image.

[0060] Furthermore, the screening process for obtaining abnormal images includes: based on the feature information matrix of the abnormal block image, determining whether the feature information matrix converted from the abnormal video frame image contains a feature similarity matrix whose similarity exceeds a set similarity; if so, marking the abnormal video frame image as an abnormal image.

[0061] Preferably, there is no limitation on the screening method for screening images containing abnormal block images from abnormal video frame images according to abnormal block image information to obtain abnormal images, and any image screening method that can realize the above screening process can be used, for example, an image screening algorithm constructed based on a convolutional neural network, an image screening algorithm constructed based on machine learning, etc.

[0062] Preferably, the preset number is greater than or equal to 50, and the specific preset number is determined according to actual needs.

[0063] Preferably, the preset abnormal information library stores multiple abnormal levels of each classification type constructed based on expert experience and historical data, as well as the membership degree values ​​corresponding to the number of abnormal images and the abnormal level, and the membership degree values ​​corresponding to the proportion of abnormal block images in abnormal images and the abnormal level.

[0064] In this embodiment, based on the proportion of abnormal block images in abnormal images in the preset abnormal information library and the membership degree value of the abnormal level, the averaging method is used to calculate the first membership value of each abnormal image; the second membership value is the membership degree value corresponding to the number of abnormal images and the abnormal level.

[0065] The present invention determines the abnormal level by presetting an abnormal information library to determine the membership degree value corresponding to the number of abnormal images and the abnormal level, the occupancy ratio of abnormal block images in each abnormal image and the membership degree value of the abnormal level, thereby improving the accuracy of determining the abnormal level.

[0066] In one embodiment of the present invention, the information analysis unit at least includes a weather analysis unit, a vehicle analysis unit, a high-speed ground analysis unit, and a surrounding environment analysis unit. The weather analysis unit is used to analyze the weather conditions in the image, including but not limited to heavy rain, rainstorm, hail, heavy snow, blizzard, etc. The vehicle analysis unit is used to analyze the vehicle status of other vehicles in the image, including but not limited to vehicle type, relative movement speed of vehicles, etc. The high-speed ground analysis unit is used to analyze the ground status of the highway, including but not limited to collapse, potholes, and spilled objects from other vehicles, etc. The surrounding environment analysis unit is used to analyze the forest status and river status beside the highway in the image.

[0067] The present invention also provides a multi-scenario monitoring and early warning system based on big data, the multi-scenario monitoring and early warning system includes a data acquisition module, a data preprocessing module, an image analysis module, an abnormality analysis module, a data storage module and an early warning module; the multi-scenario monitoring and early warning system is used to implement the monitoring and early warning method described in any one of the above embodiments through mutual cooperation between the modules.

[0068] Specifically, the data acquisition module is used to collect in real time the monitoring image information and location information obtained by the ETC video integrated device on the vehicle on the highway.

[0069] The data preprocessing module is used to preprocess the monitoring image information to obtain a preprocessed image; and convert the preprocessed image into a video frame image arranged in chronological order.

[0070] The image analysis module is used to determine whether the degree of change of the video frame image at the next moment is greater than the set degree value based on the video frame image at the previous moment; if so, cluster analysis is performed on the video frame image at the next moment to determine the division type of the video frame image at the next moment, and a feature extraction model is used to extract feature information in the video frame image at the next moment to form a feature information matrix; and according to the division type, the corresponding information analysis unit in the image analysis model is used to analyze the feature information matrix to obtain the image analysis result.

[0071] The abnormal analysis module is used to obtain abnormal video frame images of a set time period starting from a later moment when abnormal information exists in the image analysis result, and analyze the abnormal video frame images using an abnormal analysis model to obtain an abnormal analysis result.

[0072] The data storage module is used to store the input data, output data of each module, and data of the preset information library, preset abnormal information library, etc.

[0073] The warning module is used to generate warning information based on the abnormal analysis results and location information; and obtain the driver information who is on the highway within a set range and has not driven to the location information; and send warning information to the driver through the driver information.

[0074] So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.

Claims

1. A multi-scenario monitoring and early warning method based on big data, characterized in that: include: Real-time collection of monitoring image information and location information obtained by the ETC video integrated device on the highway; Preprocessing the monitoring image information to obtain a preprocessed image; and converting the preprocessed images into video frame images arranged in time sequence; Based on the video frame image at the previous moment, it is determined whether the change degree of the video frame image at the next moment is greater than the set degree value; if so, a cluster analysis is performed on the video frame image at the next moment to determine the division type of the video frame image at the next moment, and a feature extraction model is used to extract feature information in the video frame image at the next moment to form a feature information matrix; According to the division type, the corresponding information analysis unit in the image analysis model is used to analyze the feature information matrix to obtain the image analysis result; When there is abnormal information in the image analysis result, an abnormal video frame image of a set time period starting from the next moment is obtained, and the abnormal video frame image is analyzed using an abnormal analysis model to obtain an abnormal analysis result; Generate warning information based on abnormal analysis results and location information; and obtain driver information that is on the highway within a set range and has not driven to the location information; and send warning information to the driver through the driver information.

2. The monitoring and early warning method according to claim 1, characterized in that: The method of judging whether the degree of change of the video frame image at the next moment is greater than a set degree value based on the video frame image at the previous moment includes: converting the video frame image into a pixel information set of the corresponding channel through the three color channels of R, G, and B, and storing them in chronological order; obtaining the pixel information set at the previous moment and the pixel information set at the next moment; calculating the degree of change between the pixel information set at the next moment and the pixel information set at the previous moment, and the degree of change value The calculation formula is: ， in, is the i-th pixel in the pixel information set of the R channel at the next moment, is the i-th pixel in the pixel information set of the R channel at the previous moment, is the i-th pixel in the pixel information set of the G channel at the next moment, is the i-th pixel in the pixel information set of the G channel at the previous moment, is the i-th pixel in the pixel information set of the B channel at the next moment, is the i-th pixel in the pixel information set of the B channel at the previous moment; judge the degree of change Is it greater than the set value? 3. The monitoring and early warning method according to claim 1, characterized in that: The clustering analysis is performed on the video frame image at the next moment to determine the division type of the video frame image at the next moment, including: converting the video frame image at the next moment into an initial clustering data set through a preset converter; and analyzing the initial clustering data set using a clustering model constructed based on historical image data to obtain the division type.

4. The monitoring and early warning method according to claim 1, characterized in that: The extraction process of the feature extraction model includes: dividing the video frame image at the next moment into multiple irregular block images; according to the division type, screening out the block images that meet the corresponding division type from each block image to obtain multiple key analysis images; using a feature extractor to extract feature information of each key analysis image to form multiple feature information matrices.

5. The monitoring and early warning method according to claim 4, characterized in that: The processing process of the image analysis model includes: according to the division type, each feature information matrix is ​​input into the corresponding information analysis unit; each information analysis unit judges whether the feature information matrix meets the set standards of the preset feature information library according to its own preset feature information library, and when it does not meet the set standards, an abnormal block image is obtained, and an image analysis result containing abnormal block image information is obtained.

6. The monitoring and early warning method according to claim 5, characterized in that: The setting standard of the preset feature information library includes a standard setting interval of each element in the feature information matrix and a setting variation interval between the element and surrounding elements.

7. The monitoring and early warning method according to claim 5, characterized in that: The analysis process of the abnormal analysis model includes: according to the abnormal block image information, images containing abnormal block images are screened from abnormal video frame images to obtain abnormal images; when the number of abnormal images exceeds a preset number, each abnormal image is analyzed based on a preset abnormal information library corresponding to the division type, and a first membership value of each abnormal image and a second membership value corresponding to the number of abnormal images are determined; the first membership value and the second membership value are processed by a weighted average method, and an average membership value is calculated, and the level corresponding to the average membership value is used as the abnormal level; and an abnormal analysis result is formed according to the division type, the abnormal level and each abnormal image.

8. The monitoring and early warning method according to claim 7, characterized in that: The screening process for obtaining abnormal images includes: based on the feature information matrix of the abnormal block image, determining whether the feature information matrix converted from the abnormal video frame image contains a feature information matrix whose similarity exceeds a set similarity; if so, marking the abnormal video frame image as an abnormal image.

9. The monitoring and early warning method according to claim 1, characterized in that: The information analysis unit at least includes a weather analysis unit, a vehicle analysis unit, a high-speed ground analysis unit, and a surrounding environment analysis unit.

10. A multi-scenario monitoring and early warning system based on big data, characterized in that: The multi-scenario monitoring and early warning system includes a data acquisition module, a data preprocessing module, an image analysis module, an abnormality analysis module, a data storage module and an early warning module; the multi-scenario monitoring and early warning system is used to implement the monitoring and early warning method described in any one of claims 1-9 through mutual cooperation between the modules.

Images