Typhoon green risk prediction method, device and medium based on multi-source data

By using multi-source data processing technology and visual analysis models and large language models to generate risk prediction results, the problem of accuracy in predicting risks to green areas before typhoons is solved, and the probability of damage caused by typhoons is reduced.

CN122432785APending Publication Date: 2026-07-21GUANGZHOU SMART CITY INVESTMENT & OPERATION CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU SMART CITY INVESTMENT & OPERATION CO LTD
Filing Date
2026-05-27
Publication Date
2026-07-21

AI Technical Summary

Technical Problem

Existing technologies provide coarse-grained and qualitative risk predictions for green areas before typhoons, failing to pinpoint the exact location and appropriate countermeasures. This results in a high probability of personal injury and facility damage caused by the collapse of greening structures.

Method used

By acquiring multi-source data and performing modal unification preprocessing, visual assessment vectors and disaster impact vectors are generated using visual analysis models and large language models. Combined with a reinforcement learning fusion module, the fusion weights are dynamically adjusted, and risk prediction results are generated by comprehensively considering vegetation health and facility distance characteristics.

Benefits of technology

It has improved the accuracy and targeting of greening risk identification, and reduced the probability of personal injury and damage to important facilities caused by fallen greenery.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122432785A_ABST
    Figure CN122432785A_ABST
Patent Text Reader

Abstract

The application provides a typhoon green risk prediction method and device based on multi-source data and a medium. The method obtains multi-source typhoon green monitoring original data, and performs modal unification preprocessing on the multi-source typhoon green monitoring original data to generate a visual standardization feature vector suitable for a visual analysis model and a structured sequence feature vector suitable for a large language model. Then, the corresponding visual evaluation vector and the disaster influence vector are jointly input into a reinforcement learning fusion module, a feedback reward signal based on the vegetation health degree and the lodging influence range is taken as an optimization target, the fusion weight is dynamically adjusted, and the spatial distance features of the green objects and the surrounding sensitive facilities in the map POI data are comprehensively generated to generate a risk prediction result containing a risk level and a disposal suggestion, thereby improving the accuracy and pertinence of green risk identification before the typhoon arrives, and reducing the probability of personal injury and important facility damage caused by green lodging.
Need to check novelty before this filing date? Find Prior Art