System and method for constructing disaster target feature library and automatically extracting disaster area

By building a disaster target feature library and using remote sensing image processing technology and machine learning algorithms, the automated extraction and monitoring of various natural disasters is achieved, and the problem of insufficient efficiency and accuracy of disaster information extraction in the existing technology is solved, providing real-time Data support.

CN118916506APending Publication Date: 2024-11-08GUIZHOU TUZHI INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411106285.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing disaster information extraction methods are difficult to achieve comprehensive identification and dynamic monitoring of various natural disaster types, and the remote sensing image data is large and the characteristics are complex, resulting in insufficient efficiency and accuracy of disaster information extraction.

Method used

Build a disaster target feature library, combine remote sensing image processing technology and machine learning algorithms, design interpreted sample management, image preprocessing, disaster pattern extraction and human-computer interaction modules to realize automated extraction and monitoring of disaster pattern.

Benefits of technology

It improves the efficiency and accuracy of disaster information extraction, provides rapid response and accurate assessment of a variety of natural disasters, improves the accuracy, timeliness and comprehensiveness of disaster monitoring, and provides real-time data support for disaster management.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118916506A_ABST
    Figure CN118916506A_ABST
Patent Text Reader

Abstract

The invention discloses a system and a method for constructing a disaster target feature library and automatically extracting a disaster area, and the system is characterized in that the disaster target feature library takes the storage and management of data information as the main function, and provides support for the links of automatic sample selection, feature optimization and the like in the disaster target automatic classification process based on a high-resolution image; therefore, a foundation is laid for rapid assessment of disaster situations and other disaster reduction and prevention applications. The disaster target feature library is specially established for recognizable ground object targets which are in high-resolution remote sensing images and are possibly or damaged in natural disasters, and has the main functions of storage and management of interpretation features, ground object shapes, main attributes and other information of the ground object targets. And support and reference are provided for disaster information interpretation based on high-resolution remote sensing through sample matching, feature optimization and other methods. Development of a high-resolution remote sensing disaster target automatic identification technology is promoted, and decision support is provided for rapid response of disaster situations and other disaster prevention and reduction applications.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of geographic information data analysis, and in particular to a system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas. Background Art

[0002] With the intensification of global climate change and the advancement of urbanization, natural disasters are frequent and their impact range is expanding. Rapid response and effective management of disaster information have become important needs of today's society. Traditional disaster monitoring and assessment methods often rely on manual interpretation and on-site investigations, which are not only time-consuming and labor-intensive, but also difficult to achieve real-time monitoring and accurate assessment of large areas. In recent years, with the rapid development of remote sensing technology and the improvement of computer processing power, it has become possible to automatically extract and analyze disaster information using remote sensing images. However, most of the existing disaster information extraction methods rely on specific disaster types and scenarios.

[0003] There is a lack of comprehensive identification and dynamic monitoring capabilities for multiple disaster types. In addition, the amount of remote sensing image data is large and the features are complex, so how to achieve efficient and accurate disaster information extraction remains a challenge.

[0004] In this context, the present invention proposes a system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas, aiming to achieve rapid response and accurate assessment of various natural disasters by constructing a comprehensive disaster target feature library and adopting advanced remote sensing image processing technology and machine learning algorithms. The system and method not only improve the efficiency and accuracy of disaster information extraction, but also provide more comprehensive, intuitive and real-time data support for disaster management.

[0005] In order to solve the above problems, the applicant proposes a system and method for constructing a disaster target feature library and automatically extracting disaster-stricken areas. Summary of the invention

[0006] The purpose of the present invention is to provide a method for constructing a graph database for the full life cycle management of construction land, so as to solve the problems in the prior art.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: 1. A system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas, characterized in that it includes:

[0008] Build a disaster target feature library to store and manage various disaster map data, grid vector data and remote sensing image data;

[0009] Design an interpretation sample management module to generate and manage interpretation samples, support the drawing of disaster sample areas and the generation of feature sample libraries;

[0010] Provide an image preprocessing module for gridding remote sensing images and disaster patch data to form gridded images and vector data packages suitable for analysis;

[0011] Implement a disaster patch extraction module to automatically extract surface disaster patches, remove false patches, classify patches, and perform quality checks;

[0012] Set up a human-computer interaction module to manage extraction tasks, display extraction results, and support query and statistics of disaster spots;

[0013] The system and method use machine learning and image recognition technology, combined with remote sensing images and a disaster target feature library, to achieve rapid and automated extraction of disaster-affected areas.

[0014] Optionally, the disaster target feature library also includes multi-temporal remote sensing image data, as well as attribute data types and graphic data types of various disasters.

[0015] Optionally, the interpretation sample management module further includes sub-modules such as image cropping, label cropping, manual refinement, quality inspection and warehousing, sample training and sample management to support the whole process management from sample generation to training.

[0016] Optionally, the image preprocessing module also includes sub-modules such as preliminary inspection, data normalization and grid processing to ensure the consistency of image data and the effectiveness of grid processing.

[0017] Optionally, the disaster spot extraction module also includes sub-modules such as pseudo spot removal, disaster spot classification, results quality inspection and business post-processing to improve the accuracy and practicality of disaster spot extraction.

[0018] Optionally, the human-computer interaction module also includes sub-modules such as task management, comprehensive display and query statistics, which support the whole process control and result display of the disaster spot extraction task.

[0019] Optionally, the disaster map extraction process supports the automatic identification and extraction of various natural disasters including but not limited to earthquakes, floods, landslides / mudslides, snowstorms, droughts, forest fires and sandstorms.

[0020] Optionally, it also includes a mechanism for continuous optimization and updating of the extraction results to ensure that the system can remain efficient and accurate as technology and data develop.

[0021] Beneficial effects: The embodiments of the present invention use satellite remote sensing images with the best timeliness to extract based on gridding, quickly locate disaster-affected areas, and achieve rapid and intelligent identification of disaster spots, greatly improving the extraction efficiency of disaster-affected areas. At the same time, it solves the accuracy, timeliness and comprehensiveness of disaster monitoring, improves the dynamic management level of disaster-affected scope and disaster types, and provides powerful data support and information services for disaster relief and post-disaster rapid reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic diagram of the structure of a system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas provided by an embodiment of the present invention;

[0023] Figure 2 A schematic diagram of a system structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] The following describes the preferred embodiments of the present invention with reference to the drawings in the specification, so that the technical content is clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.

[0025] Example 1

[0026] Reference Figure 1 , which shows a schematic diagram of the structure of a system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas provided by an embodiment of the present invention, such as Figure 1 As shown, the system may include: a disaster target feature library 110, an interpretation sample management module 120, an image preprocessing module 130, a disaster spot extraction module 140 and a human-computer interaction module 150, wherein:

[0027] The disaster target feature library is configured as various disaster spot data, grid vector data and remote sensing image data;

[0028] The interpretation sample management module is configured to generate and manage interpretation samples, and draw various disaster sample areas to generate a feature sample library;

[0029] The image preprocessing module is configured to cut the remote sensing images and disaster spot data extracted from the database according to the monitoring grid range to form gridded images and vector data packets;

[0030] The disaster spot extraction module is configured to extract surface disaster spots, remove pseudo spots from the extracted results, classify spots according to disaster types, and perform quality inspection and business post-processing on the extracted results;

[0031] The human-computer interaction module is configured to manage the extraction tasks, and to comprehensively display the extracted disaster spots, as well as query statistics of the disaster spots.

[0032] The disaster target feature library 110 can be configured as various types of disaster spot data, grid vector data and remote sensing image data;

[0033] The interpretation sample management module 120 may be configured to generate and manage interpretation samples, draw various disaster sample areas to generate a feature sample library;

[0034] The image preprocessing module 130 may be configured to cut the remote sensing images and disaster spot data extracted from the database according to the monitoring grid range to form gridded images and vector data packets;

[0035] The disaster spot extraction module 140 may be configured to extract surface disaster spots, remove pseudo spots from the extracted results, classify spots according to disaster types, and perform quality inspection and business post-processing on the extracted results;

[0036] The human-computer interaction module 150 may be configured to manage the extraction tasks, and to comprehensively display the extracted disaster spots, as well as query statistics of the disaster spots.

[0037] In a specific implementation of the present invention, the disaster target feature library stores various types of disaster map data, square kilometer grid vector data and remote sensing image data, the data types include attribute data types and graphic data types, and the remote sensing image data includes multi-temporal remote sensing images covering the monitoring area.

[0038] In another specific implementation of the present invention, the interpretation sample management module includes: the interpretation sample management module includes: an image cutting submodule, a label cutting submodule, a manual fine-tuning submodule, a quality inspection warehousing submodule, a sample training submodule and a sample management submodule, wherein:

[0039] The image cutting submodule is configured to cut the sample area image into 1024*1024 pixel images that meet the computer recognition requirements according to square kilometer grids;

[0040] The label cutting submodule is configured to cut the vector layer of the disaster map coverage status into square kilometer grids to form sample labels;

[0041] The manual refinement submodule is configured to edit the automatically cut sample labels;

[0042] The quality inspection and warehousing submodule is configured to inspect and store the cropped images and sample labels;

[0043] The sample training submodule is configured to interpret sample training, including starting, pausing, and ending a training task, and checking the progress of a training task;

[0044] The sample management submodule is configured to define and manage sample categories that can be identified by the system, and to add, modify, and delete samples, and select sample areas to participate in training.

[0045] In another specific implementation of the present invention, the image preprocessing module includes: a preliminary inspection submodule, a data normalization submodule and a grid processing submodule, wherein:

[0046] The preliminary inspection submodule is configured to extract image data from the remote sensing big database according to the monitoring grid and perform preliminary inspection of the coordinate system, projection mode and data format;

[0047] The data normalization submodule is configured to perform attribute field normalization processing on various types of disaster map data;

[0048] The grid processing submodule is configured to cut the remote sensing images and various disaster vector data extracted from the database according to the monitoring grid range to form grid images and vector data packets.

[0049] In another specific implementation of the present invention, the disaster spot extraction module includes: an element extraction submodule, a pseudo spot removal submodule, a disaster spot classification submodule, a result quality inspection submodule and a business post-processing submodule, wherein:

[0050] The element extraction submodule is configured to extract disaster sample features and perform remote sensing image recognition, and obtain corresponding recognition results;

[0051] The pseudo-spot removal submodule is configured to automatically screen the extracted disaster spots according to the characteristics of various disaster spots and remove pseudo-spots;

[0052] The disaster spot classification submodule is configured to classify the non-eliminated spots;

[0053] The product quality inspection submodule is configured to perform attribute value and topology quality inspection on the classified images;

[0054] The business post-processing submodule is configured to perform disaster spot information supplementation and business information supplementation processing.

[0055] In another specific implementation of the present invention, the human-computer interaction module includes: a task management submodule, a comprehensive display submodule and a query statistics submodule, wherein:

[0056] The task management submodule is configured to control the disaster spot extraction task and send the start, pause and stop instructions of the extraction task to the spot extraction module;

[0057] The comprehensive display submodule is configured to receive the analysis results sent by the disaster spot extraction module and display them in the form of a combination of maps, pictures and attributes;

[0058] The query and statistics submodule is configured to query and count the disaster spots, and the business personnel can count the area and number of the disaster areas.

[0059] In another specific implementation of the present invention, the disaster spot extraction includes at least one of the following rules:

[0060] The identification result at the monitoring time point is an earthquake;

[0061] The identification result at the monitoring time point is flood;

[0062] The identification result at the monitoring time point is landslide / mudslide;

[0063] The identification result at the monitoring time point was a snow disaster;

[0064] The identification result at the monitoring time point is drought;

[0065] The identification result at the monitoring time point is forest fire;

[0066] The identification result at the monitoring time point was a sandstorm.

[0067] Next, combine Figure 2 The technical solution of the embodiment of the present invention is described in detail as follows.

[0068] like Figure 2 As shown, the disaster target feature library construction and disaster-affected area automatic extraction system includes a disaster target feature library, an interpretation sample management module, an image preprocessing module, a disaster spot extraction module and a human-computer interaction module.

[0069] 1. Disaster Target Feature Database

[0070] Store remote sensing image data, square kilometer grid data, and various disaster map data, whose information includes attribute data, vector graphics data, grid image pictures, etc.

[0071] 2. Interpretation Sample Management Module

[0072] Use square kilometer grids to crop various disaster spots and the image data corresponding to the spots, and select the required sample categories through manual refinement. The sample categories are divided into earthquakes, floods, landslides / mudslides, snow disasters, droughts, low temperatures, forest fires, sandstorms, and pests and diseases. The images are compared to confirm the validity of the samples, and they are put into storage after passing the inspection and trained using the sample training module.

[0073] (1) Sample management submodule

[0074] The classification that the system can identify is defined according to business needs and the degree of identification. The disaster categories are defined according to the needs of automatic extraction of disaster-stricken areas combined with image features. The explanation is as follows:

[0075] Earthquake: The earthquake range can be visually seen through images. The epicenter area is often severely damaged. As the distance from the epicenter increases, the degree of damage will decrease accordingly.

[0076] Floods: The flood range is visually detected through images. Long-term excessive rainfall leads to a large amount of water accumulation and runoff drainage, causing waterlogging and flooding of land and houses.

[0077] Landslide / mudslide: The image shows the scope of the landslide / mudslide. A large amount of landslides or mudslides will bury or destroy buildings, farmland, roads, etc.

[0078] Snow disaster: The scope of the snow disaster is visually detected through images. Due to the large-scale accumulation of snow caused by snowfall, the survival of people, livestock and wild animals in pastoral areas is seriously affected, and transportation, electricity, communication and other facilities are damaged due to the heavy snowfall.

[0079] Drought: The earthquake range is visually detected through images. There has been less precipitation for a long time, and river runoff and other water resources are in short supply, resulting in a serious lack of soil moisture and dry vegetation.

[0080] Forest fire: The image shows the scope of the forest fire. The red flames at the fire point and the thick smoke produced by the flames can be clearly seen in the image. After the fire, the image color of the location appears black;

[0081] Sandstorm: Visually, the image shows the scope of the sandstorm. In some areas, a large area of ​​earthy yellow is seen, and crop losses are serious.

[0082] Sample management mainly implements the definition, creation, editing, and deletion of samples.

[0083] (2) Sample drawing tool

[0084] The system provides sample drawing tools, the main functions of which include label cutting, image cutting, manual retouching, quality inspection and warehousing, etc. The functions are described as follows:

[0085] Label cutting: Cut various disaster map data into square kilometer grids.

[0086] Image cropping: The remote sensing image is cropped into 1024*1024 pixel images according to square kilometer grids that meet computer recognition requirements.

[0087] Manual refinement: Edit the cropped label vector patches by comparing the images of the disaster area.

[0088] Quality inspection and warehousing: Check the cut images and sample labels and put them into storage.

[0089] (3) Sample training submodule

[0090] After the samples are drawn, they need to be trained and their features extracted to form a sample library.

[0091] Sample training mainly realizes functions such as training sample selection, training task control, task monitoring, and recognition rate checking. The functions are described as follows:

[0092] Training sample selection: used to select samples to be trained from all samples in the database.

[0093] Training task control: After selecting the training samples, start the training task. If you need to stop for some reason during the training process, you can terminate the task.

[0094] Task monitoring: used for process monitoring of training tasks, showing the training progress in the form of a progress bar, and displaying information such as the start time and expected end time.

[0095] Recognition rate check: After the sample training is completed, you can manually select one or more pictures for recognition. The manual inspection result is in line with expectations. If the result does not meet the requirements, you can add or replace samples and retrain.

[0096] 3. Image Preprocessing Module

[0097] The latest images covering the monitoring area are extracted from the remote sensing database, and preliminary inspection and normalization are automatically performed. Then, the monitoring grid is extracted using the monitoring range, and the image is cropped using the extracted grid to complete the image gridding processing.

[0098] The preliminary inspection mainly completes the coordinate system, projection method, data format, etc.; the normalization processing mainly completes the normalization of the attribute fields of various disaster map data, including the unification of the value ranges of fields such as sample type, grid number, and time phase; the gridding processing completes the cutting of remote sensing images and various map data extracted from the database to form a gridded disaster map extraction module for use.

[0099] 4. Disaster Spot Extraction Module

[0100] Extract elements from the gridded images, extract disaster spots, and classify them according to the target feature index values ​​during identification. Remove pseudo spots with low target feature index values, small areas, and abnormal shapes from the extracted disaster spots, and then check the attribute values ​​and topology, and perform post-processing such as category information supplementation and business information supplementation.

[0101] The rules for mesh change extraction are as follows:

[0102] Earthquake: The earthquake range can be visually seen through images. The epicenter area is often severely damaged. As the distance from the epicenter increases, the degree of damage will decrease accordingly.

[0103] Floods: The flood range is visually detected through images. Long-term excessive rainfall leads to a large amount of water accumulation and runoff drainage, causing waterlogging and flooding of land and houses.

[0104] Landslide / mudslide: The image shows the scope of the landslide / mudslide. A large amount of landslides or mudslides will bury or destroy buildings, farmland, roads, etc.

[0105] Snow disaster: The scope of the snow disaster is visually detected through images. Due to the large-scale accumulation of snow caused by snowfall, the survival of people, livestock and wild animals in pastoral areas is seriously affected, and transportation, electricity, communication and other facilities are damaged due to the heavy snowfall.

[0106] Drought: The earthquake range is visually detected through images. There has been less precipitation for a long time, and river runoff and other water resources are in short supply, resulting in a serious lack of soil moisture and dry vegetation.

[0107] Forest fire: The image shows the scope of the forest fire. The red flames at the fire point and the thick smoke produced by the flames can be clearly seen in the image. After the fire, the image color of the location appears black;

[0108] Sandstorm: Visually, the image shows the scope of the sandstorm. In some areas, a large area of ​​earthy yellow is seen, and crop losses are serious.

[0109] 5. Human-computer interaction module

[0110] The human-computer interaction module manages the tasks, comprehensively displays the disaster spots, and performs query statistics on the changed spots.

[0111] (1) Task management submodule

[0112] Business personnel can start, stop, delete, set the interval for image extraction and other operations on the tasks in the task list. If the system cannot find the required image within the task scope, it will prompt that there is no image, prompting the business personnel to conduct on-site verification.

[0113] (2) Comprehensive display submodule

[0114] Comprehensive display is mainly used to extract the results of the map. The monitoring result information includes administrative divisions, grid numbers, disaster categories, etc. It supports map positioning, which is convenient for business personnel to intuitively view the information of the disaster-stricken areas.

[0115] (3) Query Statistics Submodule

[0116] Users can query and count data on disaster-stricken areas. Queries are mainly based on task scope, administrative divisions, disaster categories, etc., and query results are automatically counted, including dimensions such as quantity and area.

[0117] The system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas provided by the embodiments of the present invention have the following beneficial effects:

[0118] The embodiment of the present invention uses satellite remote sensing images with the best timeliness to extract based on gridding, quickly locate the disaster-stricken area, and realize rapid and intelligent identification of disaster spots, greatly improving the extraction efficiency of the disaster-stricken area. At the same time, it solves the accuracy, timeliness and comprehensiveness of disaster monitoring, improves the dynamic management level of the disaster-stricken scope and disaster type, and provides powerful data support and information services for disaster relief and post-disaster rapid reconstruction.

[0119] The specific implementation methods described in this application can enable those skilled in the art to understand this application more comprehensively, but do not limit this application in any way. Therefore, those skilled in the art should understand that the present application is still modified or replaced by equivalents; and all technical solutions and improvements that do not deviate from the spirit and technical essence of this application should be included in the scope of protection of the patent of this application. The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic features of the present invention. Therefore, no matter from which point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the attached claims rather than the above description, and it is intended to include all changes within the meaning and scope of the equivalent elements of the claims in the present invention. Any figure mark in the claim should not be regarded as limiting the claim involved.

[0120] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A system and method for constructing a disaster target feature library and automatically extracting disaster-affected areas, characterized in that: include: Build a disaster target feature library to store and manage various disaster map data, grid vector data and remote sensing image data; Design an interpretation sample management module to generate and manage interpretation samples, support the drawing of disaster sample areas and the generation of feature sample libraries; Provide an image preprocessing module for gridding remote sensing images and disaster patch data to form gridded images and vector data packages suitable for analysis; Implement a disaster patch extraction module to automatically extract surface disaster patches, remove false patches, classify patches, and perform quality checks; Set up a human-computer interaction module to manage extraction tasks, display extraction results, and support query and statistics of disaster spots; The system and method use machine learning and image recognition technology, combined with remote sensing images and a disaster target feature library, to achieve rapid and automated extraction of disaster-affected areas.

2. The system and method according to claim 1, characterized in that: The disaster target feature library also includes multi-temporal remote sensing image data, as well as attribute data types and graphic data types of various disasters.

3. The system and method according to claim 1, characterized in that: The interpretation sample management module further includes sub-modules such as image cropping, label cropping, manual refinement, quality inspection and warehousing, sample training and sample management to support the whole process management from sample generation to training.

4. The system and method according to claim 1, characterized in that: The image preprocessing module also includes submodules such as preliminary inspection, data normalization and grid processing to ensure the consistency of image data and the effectiveness of grid processing.

5. The system and method according to claim 1, characterized in that: The disaster spot extraction module also includes sub-modules such as pseudo spot removal, disaster spot classification, results quality inspection and business post-processing to improve the accuracy and practicality of disaster spot extraction.

6. The system and method according to claim 1, characterized in that: The human-computer interaction module also includes sub-modules such as task management, comprehensive display and query statistics, which support the whole process control and result display of the disaster spot extraction task.

7. The system and method according to any one of claims 1 to 6, characterized in that: The disaster map extraction process supports the automatic identification and extraction of various natural disasters including but not limited to earthquakes, floods, landslides / mudslides, snowstorms, droughts, forest fires, sandstorms, and the like.

8. The system and method according to any one of claims 1 to 7, characterized in that: It also includes a mechanism for continuous optimization and updating of extraction results to ensure that the system can remain efficient and accurate as technology and data develop.