Disaster image information extraction method and system

Abstract

The invention discloses a disaster image information extraction method and system, and belongs to the technical field of disaster emergency response, and the method comprises the steps: planning a flight path of an unmanned plane according to a three-dimensional map of a disaster area, so as to collect the original image data of the disaster area; preprocessing the original image data to obtain preprocessed image data; wherein the preprocessing comprises image correction processing, image enhancement processing and image noise reduction processing; and carrying out feature extraction on the preprocessed image data to obtain disaster features, carrying out identification and classification on the disaster features to obtain disaster category information, and carrying out storage and visual display on the disaster category information. According to the invention, timely assessment of disasters is realized, and the subsequent rescue speed is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of disaster emergency response, and particularly relates to a disaster image information extraction method and system. BACKGROUND

[0002] With the intensification of global climate change, natural disasters such as typhoons and floods occur frequently, causing serious threats to human society. Traditional disaster investigation mainly relies on manual on-site investigation and satellite remote sensing technology, but manual investigation has slow response speed, limited coverage and safety hazards, while satellite remote sensing technology has low time resolution, limited spatial resolution and is greatly affected by weather conditions, which is difficult to meet the timeliness and accuracy requirements of post-disaster emergency response.

[0003] The rapid development of unmanned aerial vehicle technology provides a new solution for disaster emergency response, but the existing unmanned aerial vehicle disaster monitoring method still has obvious deficiencies in data collection, transmission, processing and display. Therefore, it is urgent to develop a disaster information extraction method integrating multiple technologies to realize timely evaluation and rapid rescue of disasters. SUMMARY

[0004] To solve the above problems, the application provides a disaster image information extraction method and system, which realizes timely evaluation of disasters to improve the speed of subsequent rescue.

[0005] To achieve the above object, the application provides the following technical scheme: In a first aspect, the application provides a disaster image information extraction method, which comprises: planning a flight path of an unmanned aerial vehicle according to a three-dimensional map of a disaster area to collect original image data of the disaster area; preprocessing the original image data to obtain preprocessed image data; wherein the preprocessing includes image correction processing, image enhancement processing and image noise reduction processing; extracting disaster features from the preprocessed image data, identifying and classifying the disaster features to obtain disaster category information, and visually displaying the disaster category information.

[0006] Further, the flight path of the unmanned aerial vehicle is planned according to the three-dimensional map of the disaster area to collect original image data of the disaster area, which comprises: obtaining a three-dimensional map of the disaster area from a historical database, and marking a starting point and a target collection area of the unmanned aerial vehicle in the three-dimensional map; identifying obstacles in the target collection area based on the three-dimensional map to obtain obstacle information; generating an obstacle avoidance flight path of the unmanned aerial vehicle by Dijkstra algorithm combined with the obstacle information in the three-dimensional map. Control the drone to fly along an obstacle avoidance flight path and acquire raw image data of the target area through the airborne camera; During flight, the airborne camera automatically adjusts its exposure parameters, aperture, and ISO according to ambient light conditions to obtain high-resolution images. The airborne camera includes an aerial camera and a multispectral camera. The aerial camera acquires visible light images of the ground, and the multispectral camera acquires multi-band images.

[0007] Furthermore, the method also includes compressing and encrypting the original image data, and selecting a transmission method to transmit it to the backend processing center, including: The original image data is compressed, and the compressed original image data is encrypted using an encryption algorithm to obtain encrypted original image data. Based on the transmission distance between the drone and the back-end processing center, the system adaptively selects either 5G network or satellite communication as the transmission method to send the encrypted raw image data to the back-end processing center. The network quality is monitored in real time during transmission. When the 5G network signal is lower than the preset strength, it automatically switches to satellite communication or other backup networks for transmission and retransmits encrypted image data that fails to transmit. Specifically, based on the transmission distance between the drone and the back-end processing center, the system automatically selects either 5G network or satellite communication as the transmission method, including: When the transmission distance is less than or equal to a preset threshold, 5G network is selected for transmission first. When the transmission distance is greater than a preset threshold and the ground network signal strength is lower than a preset threshold, satellite communication will be automatically switched for transmission.

[0008] Furthermore, the preprocessing of the original image data to obtain preprocessed image data includes: The calibration intrinsic parameters and distortion coefficients of the airborne camera are obtained through camera calibration technology, and the original image data is processed by radial / tangential distortion correction formula to eliminate the influence of lens distortion and obtain corrected image data. Image enhancement processing is performed on the corrected image data using histogram equalization and the Laplacian operator to obtain enhanced image data; A triple filtering method is used to perform image denoising on the enhanced image data to obtain the denoised image data, i.e., the preprocessed image data; The triple filtering method includes mean filtering, median filtering and Gaussian filtering. The enhanced image is sequentially subjected to mean filtering to remove random noise, median filtering to remove salt-and-pepper noise, and Gaussian filtering to preserve edge information.

[0009] Furthermore, the step of extracting disaster features from the preprocessed image data, and identifying and classifying the disaster features to obtain disaster category information, includes: Principal component analysis and minimum noise separation transform techniques are used to extract features from the preprocessed image data to obtain disaster characteristics; A supervised classification model is trained based on historical disaster data. The supervised classification model is then used to identify the disaster features in the image, thereby identifying the local disaster-stricken areas and damaged objects in the preprocessed image data. A supervised classification model is used to classify and analyze the local disaster-stricken areas and damaged objects to obtain disaster category information; wherein the disaster category information includes the damage category of the damaged objects and disaster-related information.

[0010] Furthermore, the method also includes storing real-time disaster data and historical disaster data, including: The real-time disaster data collected during this mission will be stored in the real-time storage module, and the historical disaster data will be stored in the historical data storage module. The real-time disaster data includes the collected raw image data and the corresponding disaster category information.

[0011] Furthermore, the visualization of the disaster category information includes: On the electronic map, the disaster category information is displayed in a visual classification, and the displayed content includes at least one of the following: disaster area, facility damage, vegetation damage, landform changes, and population flow. The damage to facilities includes damage to buildings, roads, power supply, communications, and special facilities.

[0012] Secondly, embodiments of the present invention provide a disaster image information extraction system, the system being used to implement the aforementioned method, the system comprising: The data acquisition module is used to plan the flight path of the drone based on the 3D map of the disaster area in order to collect the original image data of the disaster area; The processing and analysis module is used to preprocess the original image data, extract features from the preprocessed image data to obtain disaster features, and identify and classify the disaster features to obtain disaster category information. The preprocessing includes image correction processing, image enhancement processing, and image noise reduction processing. The results display module is used to visualize the disaster category information.

[0013] Furthermore, the data acquisition module includes: The path acquisition unit is used to obtain a three-dimensional map of the disaster-stricken area from a historical database and mark the starting point and target collection area of ​​the UAV in the three-dimensional map; The path planning unit is used to identify obstacles in the target acquisition area based on the three-dimensional map, obtain obstacle information, and generate the obstacle avoidance flight path of the UAV by combining the obstacle information in the three-dimensional map with the Dijkstra algorithm. The image acquisition unit is used to acquire raw image data of the target acquisition area through an airborne camera.

[0014] Furthermore, the system includes a data transmission module for compressing and encrypting the original image data, and selecting a transmission method to transmit it to the backend processing center; The data transmission module includes a data compression and encryption unit, a 5G network connection unit, and a satellite communication connection unit. The data compression and encryption unit is used to compress the original image data and encrypt the compressed original image data using an encryption algorithm to obtain encrypted original image data. The 5G network connection unit is used to transmit encrypted raw image data to the back-end processing center via the 5G network when the transmission distance between the drone and the back-end processing center is less than or equal to a preset threshold and there is signal coverage. The satellite communication connection unit is used to transmit encrypted raw image data to the back-end processing center via satellite communication when the transmission distance between the UAV and the back-end processing center exceeds a preset threshold or when the ground network is paralyzed.

[0015] Furthermore, the processing and analysis module includes a preprocessing submodule and an information extraction submodule; The preprocessing submodule is used to perform image correction, image enhancement, and image noise reduction on the original image data. The information extraction submodule is used to extract features from the preprocessed image data to obtain disaster features, and to identify and classify the disaster features to obtain disaster category information; in, The preprocessing submodule includes an image correction unit, an image enhancement subunit, and an image noise reduction unit; The image correction unit is used to obtain the calibration intrinsic parameters and distortion coefficients of the airborne camera through camera calibration technology, and to perform image correction processing on the original image data using the radial / tangential distortion correction formula to eliminate the influence of lens distortion and obtain corrected image data. The image enhancement subunit is used to perform image enhancement processing on the corrected image data through histogram equalization and Laplacian operator to obtain enhanced image data; The image denoising unit is used to perform image denoising processing on the enhanced image data using mean filtering, median filtering and Gaussian filtering to obtain denoised image data.

[0016] Furthermore, the information extraction submodule includes: The feature extraction unit is used to extract features from the preprocessed image data using principal component analysis and minimum noise separation transform techniques to obtain disaster features. The feature recognition unit is used to train a supervised classification model based on historical disaster data, and to use the supervised classification model to perform feature recognition on the disaster features in the image, so as to identify the disaster-stricken areas and damaged objects in the preprocessed image data; The feature classification unit is used to classify and analyze the disaster-stricken area and damaged objects using a supervised classification model to obtain disaster category information; wherein the disaster category information includes disaster category and disaster level.

[0017] Furthermore, the system also includes a storage management module for storing real-time disaster data and historical disaster data; the storage management module includes a real-time data storage unit and a historical data storage unit. The real-time data storage unit is used to store the real-time disaster data collected during this mission; the real-time disaster data includes the collected raw image data and the corresponding disaster category information. The historical data storage unit is used to store historical disaster data.

[0018] Furthermore, the results display module includes: The disaster-affected area display unit is used to show the area affected by different disaster levels; The facility damage display unit is used to display different types of facility damage; wherein, the facility damage includes building damage, road damage, power damage, communication damage, and damage to special facilities; The vegetation damage display unit is used to display the extent of vegetation damage. The landform change display unit is used to showcase changes in landforms; The crowd flow display unit is used to show the crowd flow situation.

[0019] Compared with the prior art, this application has the following advantages: 1. This application constructs a complete disaster information extraction solution by integrating advanced UAV data collection technology, intelligent data processing algorithms and multi-dimensional information display methods; 2. In terms of technology, this application achieves efficient and safe image acquisition of disaster-stricken areas through intelligent path planning and multi-source camera configuration; it effectively improves the identification accuracy of disaster information by adopting targeted image preprocessing and feature extraction algorithms; and it provides intuitive and comprehensive decision support for different users through structured visualization design. 3. In terms of application effect, this application significantly improves the efficiency and accuracy of typhoon disaster emergency response; the system can quickly generate comprehensive disaster reports containing multi-dimensional information such as the affected area, facility damage, and vegetation damage, providing a scientific basis for the allocation of rescue resources, disaster assessment, and reconstruction planning; compared with traditional methods, this application greatly shortens the disaster assessment time, reduces the risk of manual investigation, and provides strong technical support for disaster emergency management.

[0020] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of a disaster image information extraction method according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of the structure of a disaster image information extraction system according to an embodiment of the present invention is shown; Figure 3 A detailed structural schematic diagram of a disaster image information extraction system according to an embodiment of the present invention is shown. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0024] To address the shortcomings of existing technologies, this invention discloses a method for extracting disaster image information, such as...Figure 1 As shown, the method includes: Step S1: Plan the flight path of the drone based on the 3D map of the disaster area to collect raw image data of the disaster area; Step S2: Preprocess the original image data to obtain preprocessed image data; wherein, the preprocessing includes image correction processing, image enhancement processing, and image noise reduction processing; Step S3: Extract features from the preprocessed image data to obtain disaster features, identify and classify the disaster features to obtain disaster category information, and store and visualize the disaster category information.

[0025] In some specific embodiments, step S1: Plan the flight path of the UAV based on the 3D map of the disaster area to collect raw image data of the disaster area, including the following: A three-dimensional map of the disaster-stricken area is obtained from a historical database. The three-dimensional map includes topographic information, building distribution information, and road network information. The starting point and target data collection area of ​​the UAV are marked on the three-dimensional map to determine the flight range and main mission area of ​​the UAV.

[0026] Based on the 3D map, obstacles within the target acquisition area are identified, and obstacle information is obtained. Using the Dijkstra algorithm and combining the obstacle information from the 3D map, an obstacle avoidance flight path for the UAV is generated. The UAV is controlled to fly along the obstacle avoidance flight path, and the onboard camera acquires raw image data of the target acquisition area, i.e., the raw image data of the disaster-stricken area.

[0027] During flight, the airborne camera automatically adjusts its exposure parameters, aperture, and ISO according to ambient light conditions to obtain high-resolution images. The airborne camera includes an high-altitude camera and a multispectral camera. The high-altitude camera acquires visible light images of the ground, and the multispectral camera acquires multi-band images (such as infrared, red-edge, red light, and other band images).

[0028] In some specific embodiments, after step S1, the method further includes compressing and encrypting the original image data, and selecting a transmission method to transmit it to the backend processing center, including the following: The original image data is compressed to reduce the bandwidth requirement, and the compressed original image data is encrypted using an encryption algorithm to obtain encrypted original image data, ensuring the confidentiality and integrity of the data during transmission.

[0029] Based on the transmission distance between the drone and the back-end processing center, the system adaptively selects either 5G network or satellite communication as the transmission method to send encrypted raw image data to the back-end processing center, ensuring communication reliability in complex post-disaster environments. During transmission, the system monitors network quality in real time, automatically switching to satellite communication or other backup networks when the 5G network signal strength is lower than a preset value, and retransmitting encrypted image data that fails to transmit.

[0030] Specifically, based on the transmission distance and network conditions between the drone and the back-end processing center, the transmission method is adaptively selected to use either 5G network or satellite communication, including: When the transmission distance is less than or equal to a preset threshold, 5G network is selected for high-speed transmission; when the transmission distance is greater than the preset threshold and the ground network signal is lower than the preset strength (i.e., the ground network signal is paralyzed), satellite communication or other backup networks are automatically switched for transmission.

[0031] In this embodiment, 5G network and satellite communication are used for data transmission. 5G network features high speed and low latency, enabling fast and stable transmission of large amounts of image data. Satellite communication offers advantages such as wide coverage and lack of terrain limitations, making it suitable for long-distance data transmission.

[0032] In some specific embodiments, step S2: preprocessing the original image data to obtain preprocessed image data; wherein, the preprocessing includes image correction processing, image enhancement processing, and image noise reduction processing, including the following: (1) Image correction processing: Lens distortion is an image deformation caused by the optical characteristics of a camera lens, and common types of distortion include radial distortion and tangential distortion.

[0033] Therefore, this application obtains the calibration intrinsic parameters and distortion coefficients of the airborne camera through camera calibration technology, and performs image correction processing on the original image data using the radial / tangential distortion correction formula to eliminate the influence of lens radial / tangential distortion and obtain corrected image data.

[0034] The radial distortion correction formula is as follows:

[0035] In the formula, (x corrected ,yc orrected) Let (x, y) represent the pixel coordinates of the image after radial distortion correction, (x, y) represent the pixel coordinates of the original image, and r represent the distance from the pixel to the image center. 2 =x 2 +y 2 k1, k2, and k3 all represent radial distortion coefficients.

[0036] The formula for tangential distortion correction is:

[0037] In the formula, (x corrected ,yc orrected) Let (x, y) represent the pixel coordinates of the image after tangential distortion correction, (x, y) represent the pixel coordinates of the original image, and r represent the distance from the pixel to the image center. 2 =x 2 +y 2 p1 and p2 both represent tangential distortion coefficients.

[0038] (2) Image enhancement processing: Histogram equalization and the Laplacian operator are used to perform contrast enhancement and sharpening on the corrected image data to obtain enhanced image data. Specifically, the gray-level histogram of the corrected image is calculated, and then the cumulative distribution function is calculated to map the gray-level values ​​of the corrected image to new gray-level values, making the histogram of the output image approximately uniformly distributed. Then, the Laplacian operator is used for image sharpening, and convolution operations are used to enhance the edge and detail information of the image.

[0039] The formula for histogram equalization is as follows: The grayscale histogram of the corrected image is calculated as follows: , In the formula, Represents grayscale level The probability of the total number of pixels in the corrected image. Indicates the gray level of the corrected image; Represents grayscale level The number of pixels, This represents the total number of pixels in the corrected image; The cumulative distribution function is calculated as follows: , In the formula, This represents the transformed grayscale value, i.e., the grayscale level in the corrected image. The new grayscale value obtained after mapping by the transformation function T(); This indicates the number of gray levels, typically 256. T represents the gray level of the corrected image, T() represents the transformation function, which acts as a mapping function to map the input gray level to the output gray level; k represents the index of the target gray level, and j represents the j-th gray level in the summation process.

[0040] The formula for the Laplace operator is as follows: , In the formula, This indicates that the Laplacian operator acts on the image function. The results are used for edge detection and image sharpening. This indicates the coordinates of the image after histogram equalization. The grayscale value at that location.

[0041] (3) The enhanced image data is subjected to image denoising processing using a triple filtering method to obtain the denoised image data, i.e., the preprocessed image data. The triple filtering method includes mean filtering, median filtering and Gaussian filtering.

[0042] Specifically, mean filtering smooths the image by calculating the average value of pixels in the neighborhood to remove random noise in the enhanced image; median filtering smooths the image by calculating the median value of pixels in the neighborhood to remove salt-and-pepper noise in the enhanced image; and Gaussian filtering uses a Gaussian function to perform a weighted average of pixels in the neighborhood, which can smooth the image while preserving edge information in the enhanced image.

[0043] In some specific embodiments, step S3 involves: extracting features from the preprocessed image data to obtain disaster features, identifying and classifying the disaster features to obtain disaster category information, and visualizing the disaster category information, including the following: (1) Feature extraction: Principal component analysis (PCA) and minimum noise separation transform (MNF) were used to extract features from the preprocessed image data to obtain disaster characteristics.

[0044] Specifically, principal component analysis is used to reduce the dimensionality of the preprocessed image to extract the main feature information and reduce data redundancy. Then, minimum noise separation transformation is used to separate noise and signal in the preprocessed image to enhance the useful information. Finally, feature extraction is performed on the preprocessed image to obtain disaster features, providing basic data for subsequent feature recognition and classification.

[0045] (2) Feature recognition: A supervised classification model is trained based on historical disaster data. This model compares extracted disaster features with historical disaster data in a database to identify localized disaster areas and damaged objects in the preprocessed image data. For example, the supervised classification model can identify building collapses, road breaks, and the extent of flooding.

[0046] The disaster category information is obtained by classifying and analyzing the local disaster-stricken area and damaged objects using a supervised classification model. The disaster category information includes the damage category of the damaged objects and disaster-related information.

[0047] The damage categories for damaged objects include minor damage, moderate damage, and severe damage.

[0048] For example, damage to buildings can be classified into different types such as roof damage, wall cracks, and overall collapse; damage to roads can be classified into different types such as pavement cracks, roadbed collapse, and bridge breakage.

[0049] (3) Visual presentation: On the electronic map, disaster category information is displayed in a visual classification, including at least one of the following: disaster area, facility damage, vegetation damage, landform changes, and pedestrian flow; wherein, the facility damage includes building damage, road damage, power damage, communication damage, and damage to special facilities.

[0050] In some specific embodiments, the method further includes storing real-time disaster data and historical disaster data, including the following: The real-time disaster data collected in this mission will be stored in the real-time storage module, and the historical disaster data will be stored in the historical data storage module for comparative analysis and training of the supervised classification model.

[0051] The real-time disaster data includes raw image data, corresponding disaster category information, route planning data, etc., providing data support for subsequent results display and analysis; at the same time, it can quickly generate disaster assessment reports, providing real-time information support for rescue operations.

[0052] The historical disaster data includes relevant data on historical typhoon disasters, specifically including past typhoon paths, affected areas, damaged objects, and rescue operation records; this relevant data on historical typhoon disasters provides important reference for subsequent disaster assessments and rescue operations.

[0053] In this embodiment, by storing historical disaster data, different typhoon events can be compared and analyzed to summarize the patterns and characteristics of disaster occurrence, providing a scientific basis for future disaster prevention and emergency response. For example, analyzing historical disaster data can reveal that certain areas are more susceptible to typhoon damage, allowing for the implementation of preventative measures in advance.

[0054] Based on the same inventive concept, embodiments of the present invention disclose a disaster image information extraction system for implementing the aforementioned method, such as... Figure 2 As shown, the system includes a data acquisition module, a processing and analysis module, and a results display module connected in sequence. The data acquisition module is used to plan the flight path of the drone based on the 3D map of the disaster area in order to collect the original image data of the disaster area; The processing and analysis module is used to preprocess the original image data, extract features from the preprocessed image data to obtain disaster features, and identify and classify the disaster features to obtain disaster category information. The preprocessing includes image correction processing, image enhancement processing, and image noise reduction processing. The results display module is used to visually display the disaster category information.

[0055] In some specific embodiments, the data acquisition module includes a path acquisition unit, a path planning unit, and an image acquisition unit. The path acquisition unit is connected to the path planning unit, and the image acquisition unit includes an airborne camera. The path acquisition unit is used to obtain a three-dimensional map of the disaster-stricken area from a historical database and mark the starting point and flight collection area of ​​the UAV in the three-dimensional map; The path planning unit is used to identify obstacles in the flight acquisition area based on the three-dimensional map, obtain obstacle information, and generate the obstacle avoidance flight path of the UAV by combining the obstacle information in the three-dimensional map with the Dijkstra algorithm. The image acquisition unit is used to acquire raw image data of the target area through an airborne camera.

[0056] The airborne camera includes an aerial camera and a multispectral camera. The aerial camera acquires visible light images of the ground, and the multispectral camera acquires multi-band images.

[0057] In some specific embodiments, such as Figure 3 As shown, the system also includes a data transmission module, and the data acquisition module is connected to the processing and analysis module through the data transmission module; The data transmission module is used to compress and encrypt the original image data, and select a transmission method to transmit it to the processing and analysis module in the background processing center.

[0058] The data transmission module includes a data compression and encryption unit, a 5G network connection unit, and a satellite communication connection unit, wherein the compression and encryption unit is connected to the 5G network connection unit and the satellite communication connection unit, respectively.

[0059] The data compression and encryption unit is used to compress the original image data and encrypt the compressed original image data using an encryption algorithm to obtain encrypted original image data.

[0060] The 5G network connection unit is used to transmit encrypted raw image data to the host device via a 5G network when the transmission distance between the drone and the back-end processing center is less than or equal to a preset threshold and there is signal coverage. The 5G network connection unit supports multiple 5G frequency bands and can select the optimal frequency band for connection based on different environmental conditions; it also has an automatic switching function, automatically switching to other frequency bands or backup networks when signal strength decreases to ensure continuous data transmission.

[0061] The satellite communication connection unit is used to transmit encrypted raw image data to the back-end processing center via satellite communication when the transmission distance between the UAV and the back-end processing center exceeds a preset threshold or when the ground network is paralyzed. The satellite communication connection unit supports multiple satellite communication protocols and can be flexibly configured according to different satellite systems. During transmission, the satellite communication connection unit automatically adjusts the transmission power and modulation / demodulation parameters to ensure stable data transmission. The satellite communication connection unit also has a data error correction function, which can automatically detect and correct erroneous data during transmission, improving data reliability.

[0062] In some specific embodiments, the processing and analysis module includes a preprocessing submodule and an information extraction submodule; The preprocessing submodule is used to perform image correction, image enhancement, and image noise reduction on the original image data. The information extraction submodule is used to extract features from the preprocessed image data to obtain disaster features, and to identify and classify the disaster features to obtain disaster category information.

[0063] In some specific embodiments, the preprocessing submodule includes an image correction unit, an image enhancement subunit, and an image noise reduction unit; The image correction unit is used to obtain the calibration intrinsic parameters and distortion coefficients of the airborne camera through camera calibration technology, and to perform image correction processing on the original image data using the radial / tangential distortion correction formula to eliminate the influence of lens distortion and obtain corrected image data. The image enhancement subunit is used to perform image enhancement processing on the corrected image data through histogram equalization and Laplacian operator to obtain enhanced image data; The image denoising unit is used to perform image denoising processing on the enhanced image data using mean filtering, median filtering and Gaussian filtering to obtain denoised image data.

[0064] In some specific embodiments, the information extraction submodule includes a feature extraction unit, a feature recognition unit, and a feature classification unit; The feature extraction unit is used to extract disaster features from the preprocessed image data using principal component analysis (PCA) and minimum noise separation transform (MNF) techniques, providing basic data for subsequent feature recognition and classification.

[0065] The feature recognition unit is used to train a supervised classification model based on historical disaster data, and to use the supervised classification model to perform feature recognition on the disaster features in the image, so as to identify the disaster-stricken areas and damaged objects in the preprocessed image data.

[0066] The feature classification unit is used to classify and analyze the disaster-stricken area and damaged objects through a supervised classification model to obtain disaster category information; wherein the disaster category information includes disaster category and disaster level, providing detailed information support for disaster assessment and rescue operations.

[0067] In some specific embodiments, the system further includes a storage management module; The storage management module is used to store real-time disaster data and historical disaster data.

[0068] The storage management module includes a real-time data storage unit and a historical data storage unit. The real-time data storage unit is used to store the real-time disaster data collected during this mission through a real-time database; the real-time disaster data includes the collected raw image data, corresponding disaster category information, and path planning data, etc. The historical data storage unit is used to store historical disaster data through a historical database for comparative analysis and to supervise the training of the classification model.

[0069] In some specific embodiments, the results display module includes a disaster area display unit, a facility damage display unit, a vegetation damage display unit, a landform change display unit, and a pedestrian flow display unit; The disaster area display unit is used to visually show the boundaries and area of ​​the disaster zone through map markings and color coding. For example, red can be used to mark severely affected areas, yellow to mark moderately affected areas, and green to mark lightly affected areas. In this way, rescue workers can quickly identify key rescue areas and allocate rescue resources rationally.

[0070] The facility damage display unit categorizes and displays different types of facility damage, including at least one of the following: building damage, road damage, power outages, communication disruptions, and damage to special facilities. This allows rescue personnel to quickly assess the risk of secondary disasters and take appropriate measures, such as establishing temporary shelters and sealing off dangerous areas.

[0071] The vegetation damage display unit uses high-resolution images and 3D models to show vegetation damage, enabling rescue personnel to quickly assess its impact on traffic and buildings and take appropriate measures. Vegetation damage includes fallen and broken trees; corresponding measures include clearing fallen trees and reinforcing damaged buildings.

[0072] The landform change display unit is used to show landform changes through high-resolution images and 3D models, enabling rescuers to quickly assess the risk of geological disasters and take corresponding measures, such as setting up warning signs and carrying out geological reinforcement.

[0073] The crowd flow display unit is used to show crowd flow through images and map annotations; it enables rescuers to quickly identify rescue targets and priorities, rationally allocate rescue resources, and improve rescue efficiency.

[0074] In some specific embodiments, the facility damage display unit includes a building damage display subunit, a road damage display subunit, a power damage display subunit, a communication damage display subunit, and a special facility display subunit; The building damage display sub-unit is used to display the extent of damage to a building through high-resolution images and 3D models; for example, to display the collapse, tilting, cracks and other damage to the building.

[0075] The road damage display sub-unit is used to show the road's traffic conditions and damage status through images and map annotations. Damage status includes, for example, displaying road and bridge fractures, collapses, water accumulation, and damage to traffic signs and traffic lights.

[0076] The power damage display subunit is used to display the damage to buildings, enabling rescuers to quickly identify buildings requiring emergency rescue and take appropriate measures.

[0077] The communication damage display subunit is used to show the extent of damage to communication facilities through images and map annotations, enabling rescue personnel to quickly identify communication facilities that need to be repaired first and to rationally arrange the location of temporary communication facilities.

[0078] The special facilities display sub-unit showcases the damage to special facilities, enabling rescue personnel to quickly assess the risk of secondary disasters and take appropriate measures. Special facilities include public facilities such as schools, hospitals, and shelters, as well as chemical plants, gas stations, and oil depots. Corresponding measures include establishing temporary shelters and sealing off dangerous areas.

[0079] Regarding the system in the above embodiments, the specific manner in which each unit module performs operations has been described in detail in the embodiments related to the method, and will not be elaborated here.

[0080] The entire information extraction process is as follows: First, the path acquisition unit retrieves a 3D map of the area to be collected (i.e., the disaster area) from the historical data storage unit, and marks the starting point and target point (i.e., the target collection area) of the UAV on the 3D map. The path planning unit identifies obstacles in the target collection area, plans the path based on the obstacle information, and generates the obstacle avoidance flight path of the UAV. During the flight of the UAV, the image acquisition unit collects ground image information in the flight path through a high-altitude camera and a multispectral camera, and transmits it to the processing and analysis module through the data transmission module.

[0081] After data transmission is completed, the image correction unit uses camera calibration technology to obtain the camera's intrinsic parameters and distortion coefficients, and then corrects the acquired original image using a distortion correction algorithm. Next, the image enhancement unit continues to improve the overall contrast and sharpen the original image based on histogram equalization and the Laplacian operator. Finally, the image denoising unit removes noise from the original image using a triple filtering method of mean filtering, median filtering, and Gaussian filtering to improve the signal-to-noise ratio of the image, ultimately obtaining the preprocessed image.

[0082] After image processing, information extraction is performed on the preprocessed image. The feature extraction unit extracts key disaster features from the preprocessed image, the feature recognition unit identifies the extracted disaster features, and the feature classification unit compares and analyzes the extracted disaster features with historical disaster data stored in the historical database to classify and determine their different damage types, thereby obtaining disaster category information.

[0083] The classified images can be stored through a database-based storage management module. The historical data storage unit stores historical typhoon disaster data, which is convenient for comparison and analysis. The real-time data storage unit stores the real-time typhoon disaster data collected in the current session, which is convenient for subsequent display and analysis of the results.

[0084] After the data is stored, the results display module categorizes different disaster images according to disaster type, and displays images of relevant disaster types through different display units.

[0085] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A disaster image information extraction method characterized by comprising: The method comprises: According to the three-dimensional map of the disaster area, the flight path of the unmanned aerial vehicle is planned to collect the original image data of the disaster area; The original image data is preprocessed to obtain preprocessed image data; wherein the preprocessing includes image correction processing, image enhancement processing and image noise reduction processing; The disaster feature is obtained by feature extraction on the preprocessed image data, the disaster category information is obtained by identifying and classifying the disaster feature, and the disaster category information is visualized.

2. The disaster image information extraction method according to claim 1, characterized by, According to the three-dimensional map of the disaster area, the flight path of the unmanned aerial vehicle is planned to collect the original image data of the disaster area, comprising: Obtain the three-dimensional map of the disaster area from the historical database, and mark the starting point and target collection area of the unmanned aerial vehicle in the three-dimensional map; Based on the three-dimensional map, the obstacles in the target collection area are identified to obtain obstacle information; Through Dijkstra algorithm, the obstacle avoidance flight path of the unmanned aerial vehicle is generated combined with the obstacle information in the three-dimensional map; Control the unmanned aerial vehicle to fly according to the obstacle avoidance flight path, and obtain the original image data of the target collection area through the on-board camera; Wherein, in the process of flight, the on-board camera automatically adjusts the exposure parameter, aperture and ISO of the on-board camera according to the environmental light condition to obtain high-resolution images; The on-board camera includes a high-altitude camera and a multispectral camera, which collects visible light images of the ground through the high-altitude camera and collects multi-band images through the multispectral camera.

3. The disaster image information extraction method according to claim 1 or 2, characterized by, The method further comprises compressing and encrypting the original image data, and selecting a transmission mode to transmit to the background processing center, comprising: Compress the original image data, and encrypt the compressed original image data through an encryption algorithm to obtain encrypted original image data; According to the transmission distance between the unmanned aerial vehicle and the background processing center, adaptively select 5G network or satellite communication transmission mode to send the encrypted original image data to the background processing center; Real-time monitoring of network quality during transmission, automatically switching to satellite communication or other backup network for transmission when 5G network signal is lower than the preset strength, and retransmitting the encrypted image data that fails to transmit; Wherein, according to the transmission distance between the unmanned aerial vehicle and the background processing center, automatically select 5G network or satellite communication transmission mode, comprising: When the transmission distance is less than or equal to the preset threshold, preferentially select 5G network for transmission; When the transmission distance is greater than the preset threshold, and the ground network signal is lower than the preset strength, automatically switch to satellite communication for transmission.

4. The disaster image information extraction method according to claim 1 or 2, characterized by, The preprocessing of the original image data to obtain preprocessed image data comprises: Obtain the calibration internal parameters and distortion coefficients of the on-board camera through camera calibration technology, and use the radial / tangential distortion correction formula to correct the original image data to eliminate the influence of lens distortion and obtain corrected image data; Through histogram equalization and Laplace operator, the corrected image data is subjected to image enhancement processing to obtain enhanced image data; The enhanced image data is subjected to image noise reduction processing by means of triple filtering, to obtain image data after noise reduction, i.e., image data after preprocessing. The triple filtering includes mean filtering, median filtering and Gaussian filtering, and the enhanced image data is subjected to mean filtering to remove random noise, median filtering to remove salt and pepper noise, and Gaussian filtering to retain edge information.

5. The disaster image information extraction method of claim 1, wherein, The feature extraction on the image data after preprocessing obtains disaster features, and the disaster features are identified and classified to obtain disaster category information, including: The feature extraction on the image data after preprocessing is performed by means of principal component analysis and minimum noise separation transform technology, to obtain disaster features; The supervised classification model is trained according to historical disaster data, and the disaster features in the image are subjected to feature identification by means of the supervised classification model, to identify local disaster areas and damaged objects in the image data after preprocessing; The local disaster areas and damaged objects are subjected to classification analysis by means of the supervised classification model, to obtain disaster category information; the disaster category information includes damage categories of damaged objects and disaster-related information.

6. The disaster image information extraction method of claim 1, wherein, The method further includes storing real-time disaster data and historical disaster data, including: The real-time disaster data collected in the current task is stored in a real-time storage module, and the historical disaster data is stored in a historical data storage module; The real-time disaster data includes collected original image data and corresponding disaster category information.

7. The disaster image information extraction method of claim 1, wherein The visualization of the disaster category information includes: The disaster category information is subjected to visual classification display on an electronic map, and the display content includes at least one of disaster-affected areas, facility damage, vegetation damage, landform changes and human flow conditions; The facility damage includes building damage, road damage, power damage, communication damage and special facility damage.

8. A disaster image information extraction system characterized by comprising: The system is used to implement the method of any one of claims 1 to 7, and the system includes: A data collection module is configured to plan a flight path of a UAV according to a three-dimensional map of a disaster area, to collect original image data of the disaster area; A processing and analysis module is configured to preprocess the original image data, extract disaster features from the image data after preprocessing, and identify and classify the disaster features to obtain disaster category information; the preprocessing includes image correction processing, image enhancement processing and image noise reduction processing; An achievement display module is configured to visualize the disaster category information.

9. The disaster image information extraction method according to claim 8, characterized by, The data collection module includes: A path acquisition unit is configured to acquire a three-dimensional map of a disaster area from a historical database, and mark a starting point and a target collection area of a UAV in the three-dimensional map; A path planning unit is configured to identify obstacles in the target collection area based on the three-dimensional map to obtain obstacle information, and generate an obstacle-avoiding flight path of the UAV by means of Dijkstra algorithm combined with the obstacle information in the three-dimensional map; An image acquisition unit is configured to acquire original image data of the target collection area by means of an onboard camera.

10. The disaster image information extraction system according to claim 8, characterized by, The system comprises a data transmission module for compressing and encrypting the original image data and transmitting the original image data to a background processing center by selecting a transmission mode; The data transmission module comprises a data compression and encryption unit, a 5G network connection unit, and a satellite communication connection unit; The data compression and encryption unit is configured to compress the original image data and encrypt the compressed original image data by using an encryption algorithm to obtain encrypted original image data; The 5G network connection unit is configured to transmit the encrypted original image data to the background processing center by using a 5G network when the transmission distance between the unmanned aerial vehicle and the background processing center is less than or equal to a preset threshold and there is signal coverage; The satellite communication connection unit is configured to transmit the encrypted original image data to the background processing center by using satellite communication when the transmission distance between the unmanned aerial vehicle and the background processing center is greater than the preset threshold or the ground network is paralyzed.

11. The disaster image information extraction system according to claim 8, characterized by, The processing and analysis module comprises a preprocessing submodule and an information extraction submodule; The preprocessing submodule is configured to perform image correction processing, image enhancement processing, and image noise reduction processing on the original image data; The information extraction submodule is configured to extract features from the preprocessed image data to obtain disaster features, and identify and classify the disaster features to obtain disaster category information; The preprocessing submodule comprises an image correction unit, an image enhancement subunit, and an image noise reduction unit; The image correction unit is configured to obtain calibration intrinsic parameters and distortion coefficients of an onboard camera by using a camera calibration technique, and perform image correction processing on the original image data by using a radial / tangential distortion correction formula to eliminate the influence of lens distortion and obtain corrected image data; The image enhancement subunit is configured to perform image enhancement processing on the corrected image data by using histogram equalization and a Laplacian operator to obtain enhanced image data; The image noise reduction unit is configured to perform image noise reduction processing on the enhanced image data by using mean filtering, median filtering, and Gaussian filtering to obtain noise-reduced image data. The information extraction submodule comprises:

12. The disaster image information extraction system according to claim 11, characterized by, A feature extraction unit configured to extract features from the preprocessed image data by using principal component analysis and minimum noise fraction transform technology to obtain disaster features; A feature recognition unit configured to train a supervised classification model according to historical disaster data, and use the supervised classification model to recognize the disaster features in the image to identify disaster-affected areas and damaged objects in the preprocessed image data; A feature classification unit configured to classify and analyze the disaster-affected areas and damaged objects by using the supervised classification model to obtain disaster category information; wherein the disaster category information comprises a disaster category and a disaster level. The system further comprises a storage management module configured to store real-time disaster data and historical disaster data; the storage management module comprises a real-time data storage unit and a historical data storage unit; 13. The disaster image information extraction system according to claim 8, characterized by, The real-time data storage unit is configured to store real-time disaster data collected in the current task; wherein the real-time disaster data comprises collected original image data and corresponding disaster category information. ​ The historical data storage unit is configured to store historical disaster data.

14. The disaster image information extraction system according to claim 8, characterized by, The achievement display module comprises: A disaster-affected area display unit configured to display the area range of different disaster levels; A facility damage display unit configured to display different facility damage conditions, wherein the facility damage includes building damage, road damage, power damage, communication damage, and special facility damage; A vegetation damage display unit configured to display vegetation damage conditions; A landform change display unit configured to display landform change conditions; A human flow condition display unit configured to display human flow conditions.

Images