Intelligent video surveillance analysis method

By preprocessing the video stream and calculating inter-frame differences, and combining the target detection model to generate keyframes and motion trajectories, the problems of poor frame selection accuracy and weak anti-interference ability of motion detection in the existing technology are solved, and anomaly detection with high accuracy and reliability is achieved.

CN122454293APending Publication Date: 2026-07-24SUZHOU GCL NEW ENERGY OPERATION & TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
SUZHOU GCL NEW ENERGY OPERATION & TECH CO LTD
Filing Date
2026-05-28
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing video surveillance analysis methods suffer from limitations such as simplistic and crude image filtering, which easily retain invalid interference frames and lose key event frames. Furthermore, whole-frame motion detection has weak anti-interference capabilities and poor frame filtering accuracy. Behavior analysis only models simple target coordinate changes, making it difficult to adapt to complex motion patterns. Consequently, intelligent analysis and anomaly detection suffer from low accuracy and poor reliability.

Method used

The video stream is preprocessed to generate a video frame sequence. A segmentation model is used to generate a target mask. The inter-frame optical flow difference and pixel difference are calculated to filter candidate frames. The target detection model is combined to perform target detection. The target detection score is calculated and key frames are generated. The target motion trajectory and features are extracted for anomaly detection.

Benefits of technology

It achieves precise screening of key frames, avoids the loss of invalid interference frames and key event frames, improves the accuracy of motion detection and the reliability of behavior analysis, can adapt to complex motion patterns, and improves the overall accuracy and reliability of anomaly detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122454293A_ABST
    Figure CN122454293A_ABST
Patent Text Reader

Abstract

The application belongs to the technical field of video analysis and detection, and relates to an intelligent video monitoring analysis method. Video stream is preprocessed to obtain a video frame sequence, each frame of image in the video frame sequence is input into a segmentation model, and a target mask of each frame of image is output. An optical flow difference of adjacent frame images in the video frame sequence is calculated, so that a candidate mutation frame set is obtained through screening. Based on the adjacent frame images and the target masks thereof in the video frame sequence, pixel differences of the adjacent frame images are calculated, so that a motion candidate frame set is obtained through screening. Frame images in the candidate mutation frame set and the motion candidate frame set are input into a target detection model, a target category, a target frame and coordinates thereof are output, a target detection score of each frame of image is calculated, and a key frame set is obtained through screening. Based on each frame of image and the target category, the target frame and the coordinates thereof in the key frame set, a motion trajectory of each target is obtained, a motion feature of each target is extracted and classified, so that the behavior of each target is abnormally detected.
Need to check novelty before this filing date? Find Prior Art