Campus security video inspection early warning method and system of multi-modal large model

By combining multimodal large models and thinking chain technology, intelligent inspection and hierarchical early warning of campus security have been realized, which solves the problem of time-consuming and labor-intensive traditional inspection methods and improves the accuracy of abnormal behavior identification.

CN120260245BActive Publication Date: 2026-06-09HUNAN HAND IN HAND INFORMATION TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HUNAN HAND IN HAND INFORMATION TECH CO LTD
Filing Date
2025-04-08
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Traditional campus security patrol methods are time-consuming and labor-intensive, making it difficult to detect security risks in a timely manner and unable to achieve real-time monitoring and accurate identification of abnormal behavior.

Method used

By employing a multimodal large model combined with the thinking chain technology, video information is collected through sensing devices to generate a training sample dataset. Cross-modal feature extraction and feedback iterative optimization are then performed to generate thinking chain prompt templates, enabling hierarchical early warning for real-time video data.

Benefits of technology

It has achieved intelligent and automated inspection and hierarchical early warning for campus security, and has a higher ability to accurately identify abnormal behavior.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120260245B_ABST
    Figure CN120260245B_ABST
Patent Text Reader

Abstract

The application provides a campus security video inspection early warning method and system of a multi-modal large model, and relates to the technical field of campus security. Video information corresponding to a security inspection object is collected by a preset sensor device in a target area; a preset processing condition is obtained, and the video information is processed to generate a training sample data set; a preset multi-modal large model is used in combination with the training sample data set and a thinking chain technology for optimization training, a thinking chain prompt is generated through scene description and security personnel experience, and a safety risk behavior answer fed back by the preset multi-modal large model is supplemented; based on the optimized preset multi-modal large model, an inspection mode is set, real-time video data is analyzed, and early warning is output according to the possibility of abnormal behavior classification. The application realizes intelligent and automatic inspection and hierarchical early warning of campus security, and has more accurate generalization recognition capability for abnormal behavior.
Need to check novelty before this filing date? Find Prior Art