Video event autonomous retrieval and intelligent decision system based on visual-linguistic alignment

By employing noise reduction processing, frame rate unification, cross-modal alignment, and intelligent decision-making, the challenge of aligning visual features with natural language in video retrieval systems has been solved. This enables seamless stitching of multi-camera videos and autonomous decision-making, thereby improving retrieval accuracy and decision-making efficiency.

CN122173680APending Publication Date: 2026-06-09SHANGHAI SHENSHIXIN TECH CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN Β· China
Patent Type
Applications(China)
Current Assignee / Owner
SHANGHAI SHENSHIXIN TECH CO LTD
Filing Date
2026-02-10
Publication Date
2026-06-09

AI Technical Summary

Technical Problem

Existing video retrieval systems cannot effectively align visual features with natural language queries, resulting in low retrieval accuracy, difficulty in seamlessly stitching multi-camera videos, and a lack of autonomous decision-making capabilities, which affects decision-making efficiency and accuracy.

Method used

The data acquisition unit performs noise reduction and frame rate unification to generate standardized video data; the cross-modal alignment unit performs visual-language alignment to generate structured semantic data; and the intelligent decision-making unit automatically stitches together multi-camera videos to generate a complete chain of evidence and annotates keyframes.

Benefits of technology

It enables standardized acquisition of multi-camera video and natural language queries, improves retrieval accuracy and decision-making efficiency, and provides coherent and intuitive decision-making basis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122173680A_ABST
    Figure CN122173680A_ABST
Patent Text Reader

Abstract

The present application relates to a video event autonomous retrieval and intelligent decision system based on visual-linguistic alignment. Among them, the system comprises: a data acquisition unit, a data processing unit, a cross-modal alignment unit, an intelligent decision unit; the data acquisition unit is used for acquiring first video data and second language data; through corresponding processing, standardized video data is obtained; the obtained data is processed in corresponding format, and a cross-modal alignment data set is obtained; using a corresponding feature coding model, a third semantic vector and a fourth visual vector are obtained; the matching degree score is obtained by calculating the cross-modal similarity, the video segment of the corresponding visual feature is screened, and the visual-linguistic alignment processing is carried out, and the processing result is obtained; by obtaining a complete evidence chain video, picture recognition and labeling are carried out by using a frame detection mechanism, based on the labeled evidence chain video and the matching degree score, a retrieval result report is generated; the fusion processing and intelligent analysis of video and language data are realized.
Need to check novelty before this filing date? Find Prior Art