Intelligent identification and evaluation technology and GIS map display method for tornado ground disasters

Abstract

The invention relates to an intelligent identification and evaluation technology for tornado ground disasters, and the technology comprises a data collection and preprocessing module which collects image data and text data of tornado ground disasters from a plurality of data sources, carries out the preprocessing and cleaning of the data, and generates a structured data set; the multi-modal disaster assessment model is based on a VLLM and an advanced visual identification technology, adopts a double-flow network structure and comprises an image feature extraction branch and a text processing branch, the image feature extraction branch extracts key disaster features from a high-resolution image, and the text processing branch deeply analyzes text information related to disaster grade assessment; and the interactive learning mechanism allows the multi-modal disaster assessment model to request feedback to a user after preliminary analysis, and performs self-adjustment and optimization according to the feedback. The invention further provides a map display method. According to the method, the problems of data scarcity, image diversity, low damage evaluation efficiency, rapid processing requirements and model generalization ability can be effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of natural disaster monitoring and assessment, and particularly to an intelligent recognition and assessment technology for tornado ground disaster conditions and a GIS map display method. Background Art

[0002] As an extremely destructive natural disaster, a tornado is often accompanied by strong rotating airflows and can cause serious property losses and casualties in a short period of time. After a tornado occurs, the disaster scene is easily cleaned up quickly. In order to effectively and quickly conduct a comprehensive assessment and grading of the tornado disaster afterwards, it is particularly important to carry out rapid and accurate disaster condition recognition and assessment.

[0003] Traditional methods for tornado ground disaster condition recognition and assessment mainly involve manually analyzing image and video data of the disaster situation at the scene based on domestic and foreign tornado intensity grade standards. This method is not only time-consuming and laborious, but also limited by personnel experience and judgment ability, making it difficult to ensure the accuracy and comprehensiveness of recognition. With the continuous development of information technology, especially the increasing maturity of artificial intelligence and geographic information system (GIS) technology, new possibilities have been provided for the intelligent recognition of tornado ground disaster conditions.

[0004] Although there are currently some disaster monitoring systems based on image recognition and GIS technology, these systems still have some deficiencies in the recognition of tornado ground disaster conditions. Specifically, the existing technologies mainly have the following problems: 1. Low data processing efficiency: Traditional image recognition methods often face problems such as large computational amounts and slow processing speeds when dealing with high-resolution images, making it difficult to meet real-time requirements. 2. Inaccurate feature extraction: Due to the complexity and diversity of tornado disasters, existing methods are difficult to accurately extract key features directly related to the tornado disaster damage level from images. 3. Lack of multi-modal information fusion: Existing disaster monitoring systems often only rely on image data and ignore text information related to disasters, such as descriptive texts on the damage degree and damage level of disaster indicators in domestic and foreign tornado intensity grade standards, which limits the system's comprehensive understanding of the disaster situation. 4. Single map display function: Existing GIS systems often can only provide simple geographical location markings when displaying disaster information, lacking intuitive and comprehensive display and assessment functions for disaster-affected areas.

[0005] Existing tornado ground disaster condition recognition technologies still have deficiencies in aspects such as data processing efficiency, feature extraction accuracy, multi-modal information fusion, and map display function, and a more efficient, accurate, and comprehensive intelligent recognition technology and map display method are needed to cope with the challenges of tornado disasters. Summary of the Invention

[0006] To solve the problems of low data processing efficiency, inaccurate feature extraction, lack of multi-modal information fusion, and single map display function in the existing technology for tornado ground disaster situation recognition, the present invention provides an intelligent recognition and evaluation technology for tornado ground disasters and a GIS map display method.

[0007] The specific content of the present invention is as follows: An intelligent recognition and evaluation technology for tornado ground disasters, including: A data collection and preprocessing module, which collects image data and text data of tornado ground disasters from multiple data sources, preprocesses and cleans the data, and generates a structured data set; A multi-modal disaster assessment model, based on VLLM and advanced visual recognition technology, adopting a two-stream network structure, including an image feature extraction branch and a text processing branch. The image feature extraction branch extracts key disaster features from high-resolution images, and the text processing branch deeply analyzes text information related to disaster level assessment; An interactive feedback learning mechanism, which allows the multi-modal disaster assessment model to request feedback from the user after preliminary analysis and self-adjust and optimize according to the feedback.

[0008] Furthermore, the data sources of the data collection and preprocessing module include: image materials and text descriptions of tornado disasters from meteorological agencies, real-time disaster situation image data obtained by drones taking high-resolution images of tornado occurrence areas, public feedback and on-site description texts related to tornado disasters in social media or news reports; The preprocessing of the data includes: cleaning the collected image data to remove blurred, duplicate or irrelevant images, and performing word segmentation and stop word removal operations on the text data.

[0009] Furthermore, the data collection and preprocessing module includes a data annotation step, accurately annotating the preprocessed image data, and establishing a structured database for classified storage and backup.

[0010] Furthermore, the multi-modal disaster assessment model further includes a multi-modal fusion and reasoning module. The multi-modal fusion and reasoning module fuses the outputs of the image feature extraction branch and the text processing branch, realizes deep alignment and weighting of image and text information through deep learning algorithms, and uses the fused multi-modal features to comprehensively evaluate the tornado ground disaster situation, and obtains disaster impact range and disaster situation indicator level assessment information.

[0011] Furthermore, the image feature extraction branch adopts a hierarchical or multi-scale visual model, a spatial attention mechanism, drone perspective feature learning, domain adaptation and transfer learning technologies.

[0012] Further, the text processing branch deeply understands the key disaster information in the descriptions of meteorological experts by embedding a language model that describes the damage degree and destruction level of disaster indicators.

[0013] The present invention also provides a GIS map display method, which adopts any one of the above intelligent recognition and evaluation technologies for tornado ground disaster situations, including: The data collection and preprocessing module collects the post-disaster orthophoto map and outputs the orthophoto sheet data. After being processed by the multi-modal disaster evaluation model, the disaster analysis result is output and formatted into GIS data for output. The formatted GIS data is displayed on the GIS through the GIS map program, and the disaster evaluation result is displayed in the form of a map. The disaster-affected area, disaster severity, disaster intensity, and the distribution of disaster indicators are marked on the map.

[0014] Further, it also includes a dynamic update module, which receives new disaster data in real time and updates the disaster evaluation result and map display.

[0015] Further, it also includes a user interaction interface. The user inputs query conditions through the interface to obtain a customized disaster evaluation map, and the query conditions include disaster type, disaster intensity, time range, or geographical location.

[0016] Further, it also includes a disaster warning module. When the disaster evaluation result reaches a preset threshold, the warning mechanism is automatically triggered, and warning information is sent to relevant departments and personnel through SMS, email, or APP push.

[0017] The multi-modal disaster evaluation model of the present invention adopts a two-stream network structure, which processes image data and text information related to disasters respectively, and establishes a multi-modal disaster evaluation model based on the Vision-Language Large Model (VLLM) and advanced visual recognition technology, so as to accurately and quickly evaluate the ground disaster situations caused by extreme disaster weather events such as tornadoes, and can effectively solve problems including data scarcity, image diversity, low damage assessment efficiency, rapid processing requirements, and model generalization ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The following further clarifies the specific embodiments of the present invention with reference to the drawings.

[0019] Figure 1 It is a schematic diagram of the intelligent recognition and evaluation technology for tornado ground disaster situations of the present invention; Figure 2 It is a schematic diagram of the map display method of the present invention; Figure 3 It is an example of the map display interface of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] Combined with Figure 1 Figure 1 , the present invention discloses an intelligent recognition and evaluation technology for tornado ground disaster situations. This intelligent recognition and evaluation technology is implemented based on a data layer, a two-stream network structure, a fusion layer, and an optimization mechanism. Among them, the data layer is a data collection and preprocessing module, the two-stream network structure is a multi-modal disaster evaluation model, the fusion layer is a multi-modal fusion and reasoning module, and the optimization mechanism is an interactive feedback learning mechanism.

[0021] I. Data Collection and Preprocessing Module The data collection and preprocessing module is used to collect image data and text data of tornado ground disaster situations from multiple data sources, and preprocess and clean the data to generate a structured data set. Among them: The data sources for data collection include: obtaining historical data of tornado disasters from authoritative meteorological agencies, including image materials and text descriptions; using drones to take high-resolution images of the tornado occurrence area to collect real-time disaster situation image data; collecting public feedback and on-site description texts related to tornado disasters from channels such as social media and news reports.

[0022] Data preprocessing includes: cleaning the collected image data to remove blurred, repeated, or irrelevant images; performing preprocessing operations such as word segmentation and stop word removal on the text data to improve the accuracy of subsequent text processing.

[0023] It also includes a data annotation step. A professional team precisely annotates the preprocessed image data, and establishes a structured database containing tornado disaster characteristics for classified storage and backup.

[0024] The data collection and preprocessing module can improve the model training efficiency through the cleaning, annotation, and structured storage of multi-source data (such as drones, satellites, social media, etc.).

[0025] II. Multi-modal Disaster Evaluation Model The multi-modal disaster evaluation model is based on a vision-language large model (VLLM) and advanced visual recognition technology, and adopts a two-stream network structure, including an image feature extraction branch and a text processing branch. Among them, The image feature extraction branch uses advanced visual recognition technologies such as convolutional neural networks (CNNs) to extract key disaster intensity level features from high-resolution images taken by drones, such as fallen and broken trees, damaged buildings, etc.

[0026] The image feature extraction branch uses hierarchical or multi-scale visual models, spatial attention mechanisms, drone perspective feature learning technologies, domain adaptation, and transfer learning technologies to improve the recognition accuracy of tornado disaster characteristics. The introduction of the spatial attention mechanism enables the model to focus on important regions in the image and improve the accuracy of feature extraction.

[0027] The text processing branch uses natural language processing technology (NLP) to deeply analyze the text data related to tornado disasters and extract the key information describing the disaster situation. By embedding the domestic and foreign tornado intensity level standards, expert knowledge, and context-aware language models, the model can understand the professional terms and complex situations in the descriptions of meteorological experts.

[0028] The multi-modal disaster assessment model also includes a multi-modal fusion and reasoning module, which fuses the outputs of the image feature extraction branch and the text processing branch, and realizes the deep alignment and weighting of image and text information through deep learning algorithms; uses the fused multi-modal features to comprehensively evaluate the ground disaster situation of tornadoes and obtain key information such as the disaster impact range and the assessment of the level of disaster indicators.

[0029] The multi-modal disaster assessment model uses a large vision-language model to integrate visual and language semantic information and enhance the relevance between disaster feature extraction and text description. Adopting a two-stream network structure, it can process image data and text information simultaneously, extract key disaster features through the image feature extraction branch and the text processing branch respectively, and realize the deep alignment and weighting of image and text information through the multi-modal fusion and reasoning module, thereby improving the accuracy and comprehensiveness of disaster recognition.

[0030] III. Interactive feedback learning mechanism This mechanism includes preliminary model analysis, user feedback collection, and model self-adjustment: The multi-modal disaster assessment model conducts preliminary analysis on the collected image and text data to generate disaster assessment results; displays the preliminary analysis results to users (such as meteorological experts) and collects their feedback on the assessment results; adjusts and optimizes the model parameters according to the user feedback to improve the accuracy and comprehensiveness of disaster recognition, and realizes dynamic learning and continuous improvement.

[0031] The interactive feedback learning mechanism can drive the dynamic optimization of the model through user feedback and improve the assessment accuracy through iterative adjustment.

[0032] The intelligent recognition and assessment technology for tornado ground disaster situations of the present invention can process image and text information simultaneously through the construction of a multi-modal disaster assessment model, and realize the comprehensive and accurate recognition of tornado ground disaster situations. By adopting advanced image recognition technology and natural language processing technology, the present invention can significantly improve the data processing efficiency and meet the real-time requirements.

[0033] Combined with Figure 2 and Figure 3, based on the intelligent recognition and evaluation technology for tornado ground disaster situations, the present invention also provides a GIS map display method, including GIS integration and map display. Specifically, the data collection and preprocessing module collects post-disaster orthophoto maps and outputs orthophoto mosaic data, which is processed by a multi-modal disaster assessment model (after being trained with training data and an AI model) to output disaster analysis results, and is output in the form of formatted GIS data; the formatted GIS data is displayed on the GIS through a GIS map program, and the disaster assessment results are presented in the form of a map; the disaster-affected areas, the distribution of disaster indicators, the severity of the disaster, and the intensity of the disaster are marked on the map.

[0034] The map display function provides a dynamic update mechanism. The dynamic update module receives new disaster data in real time, updates the disaster assessment results and the map display, ensuring the timeliness and accuracy of the information. A user interaction interface is designed, and users can input query conditions through the interface to obtain customized disaster assessment maps. The query conditions include disaster type, disaster intensity, time range, or geographical location.

[0035] It also includes a disaster warning module. When the disaster assessment results reach a preset threshold, the warning mechanism is automatically triggered, and warning information is sent to relevant departments and personnel through SMS, email, or APP push.

[0036] Figure 3 It shows an example of the map display interface generated under the GIS integration module of the present invention. As can be seen from the figure, information such as the specific affected areas of the tornado disaster, the severity of the disaster, and the predicted path is presented. Through intuitive graphical display, it helps users quickly understand the disaster situation.

[0037] The GIS map display method of the present invention can intuitively and comprehensively display the disaster-affected areas through GIS integration and the dynamic update mechanism, providing strong support for disaster emergency response and rescue decision-making. By constructing the disaster warning module, the present invention can send warning information in a timely manner before the disaster occurs. Combining the dynamic map and warning push, the emergency response time can be shortened by more than 30%, supporting the precise dispatching of rescue resources.

[0038] The intelligent recognition and evaluation technology for tornado ground disaster situations and the map display method of this application, through the "multi-modal AI + dynamic GIS + human-machine collaboration" trinity architecture, can intelligently recognize and evaluate tornado disaster information, reduce labor costs, achieve precise and timely response to disasters, and are especially suitable for rapid post-disaster assessment in remote areas. The solution of this application can also be extended to the assessment of natural disasters such as hurricanes and earthquakes. Through modular design, it can quickly adapt to different disaster types, and through an interactive mechanism, it can continuously absorb expert experience to improve the accuracy of the model, providing an efficient, precise, and scalable solution for disaster emergency management.

[0039] Numerous specific details have been set forth in the foregoing description in order to provide a thorough understanding of the present invention. However, the above description is merely a preferred embodiment of the present invention, and the present invention can be implemented in many other ways different from those described herein. Therefore, the present invention is not limited by the specific embodiments disclosed above. At the same time, any person skilled in the art can make many possible changes and modifications to the technical solution of the present invention, or modify it into an equivalent embodiment with equivalent changes, without departing from the scope of the technical solution of the present invention. All simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An intelligent identification and assessment technology for tornado ground disasters, characterized by: include: A data collection and preprocessing module collects image data and text data of tornado ground disasters from multiple data sources, and preprocesses and cleans the data to generate a structured data set; The multimodal disaster assessment model, based on VLLM and advanced visual recognition technology, adopts a two-stream network structure, including an image feature extraction branch and a text processing branch. The image feature extraction branch extracts key disaster features from high-resolution images, and the text processing branch deeply analyzes text information related to disaster level assessment; An interactive feedback learning mechanism allows the multimodal disaster assessment model to request feedback from users after preliminary analysis and to self-adjust and optimize based on the feedback.

2. The intelligent identification and assessment technology for tornado ground disaster according to claim 1 is characterized by: The data sources of the data collection and preprocessing module include: historical image data and text descriptions of tornado disasters from meteorological agencies, real-time disaster image data obtained by drones taking high-resolution images of tornado-affected areas, and public feedback and on-site description texts related to tornado disasters from social media or news reports; Data preprocessing includes: cleaning the collected image data, removing fuzzy, repeated or irrelevant images, and performing word segmentation and stop word removal on text data.

3. The intelligent identification and assessment technology for tornado ground disaster according to claim 2 is characterized by: The data collection and preprocessing module also includes a data labeling step to accurately label the preprocessed image data and establish a structured database for classified storage and backup.

4. The intelligent identification and assessment technology for tornado ground disaster according to claim 1 is characterized by: The multimodal disaster assessment model also includes a multimodal fusion and reasoning module, which fuses the outputs of the image feature extraction branch and the text processing branch, deeply aligns and weights the image and text information through a deep learning algorithm, and uses the fused multimodal features to comprehensively assess the tornado ground disaster, thereby obtaining disaster impact range and disaster indicator level assessment information.

5. The intelligent identification and assessment technology for tornado ground disaster according to claim 1 is characterized by: The image feature extraction branch adopts hierarchical or multi-scale visual models, spatial attention mechanism, drone view feature learning, domain adaptation and transfer learning techniques.

6. The intelligent identification and assessment technology for tornado ground disaster according to claim 1 is characterized by: The text processing branch deeply understands key disaster information by embedding a language model describing the damage extent and destruction level of disaster indicators.

7. A GIS map display method, characterized in that: The intelligent identification and assessment technology for tornado ground disasters as claimed in any one of claims 1 to 6 is adopted, comprising: a data collection and preprocessing module collects post-disaster orthophoto maps and outputs orthophoto segmented data, and outputs disaster analysis results after being processed by a multimodal disaster assessment model to format GIS data output; The formatted GIS data is displayed on the GIS through the GIS mapping program, and the disaster assessment results are presented in the form of maps; Mark the disaster-affected areas, disaster severity, disaster intensity and distribution of disaster indicators on the map.

8. The GIS map display method according to claim 7, characterized in that: It also includes a dynamic update module, which receives new disaster data in real time and updates disaster assessment results and map displays.

9. The GIS map display method according to claim 7, characterized in that: It also includes a user interaction interface, through which the user enters query conditions to obtain a customized disaster assessment map, the query conditions including disaster type, disaster intensity, time range or geographical location.

10. The GIS map display method according to claim 7, characterized in that: It also includes a disaster warning module. When the disaster assessment results reach the preset threshold, the warning mechanism is automatically triggered, and warning information is sent to relevant departments and personnel via SMS, email or APP push.

Images