First view angle video key frame extraction method and system based on job semantics

By using a keyframe extraction method based on job semantics and leveraging IMU sensors and human pose estimation, the jitter and compression issues of first-person perspective videos at power distribution network construction sites are resolved. This achieves efficient video processing and low-latency keyframe extraction, making it suitable for edge-end AR devices.

CN122457798APending Publication Date: 2026-07-24ELECTRIC POWER RES INST STATE GRID SHANXI ELECTRIC POWER
7 Cites 0 Cited by

Patent Information

Application Number
CN202610448497.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-04-07
Publication Date
2026-07-24

AI Technical Summary

Technical Problem

Existing technologies suffer from large amounts of first-person perspective video data at power distribution network construction sites, severe jitter, inaccurate traditional compression methods, and limited edge computing capabilities of AR devices, resulting in high costs, significant latency, and incomplete work content.

Method used

A keyframe extraction method based on job semantics is adopted. Motion decluttering is performed using IMU sensor data, combined with human pose estimation and adaptive compression, to extract keyframes containing job actions and generate a simplified feature data package.

Benefits of technology

It enables real-time processing at the edge of AR devices, reduces network bandwidth requirements, improves the integrity and compression ratio of job content, and has a latency of less than 200ms, making it suitable for edge deployment.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention provides a first view angle video key frame extraction method and system based on job semantics, and belongs to the technical field of computer vision. The method comprises the following steps: acquiring a first view angle video stream and IMU sensor data collected by AR equipment; motion jitter removal preprocessing is carried out based on IMU data; human body key points are extracted through a lightweight human body posture estimation model, and an operation action significance score is calculated; detecting action starting and ending points based on a saliency curve, and extracting a key frame; carrying out adaptive compression on the key frame to generate a simplified feature data packet; according to the method, key frame extraction from'picture change 'to'operation semantics' is realized according to the characteristics of the first-view-angle construction operation video, the compression ratio of 10: 1 or above is achieved while the operation action retention rate is ensured, and the method can be applied to intelligent monitoring scenes of power distribution network construction sites and has high practicability. The problem of video transmission under the condition that the network bandwidth of a distribution network site is limited is effectively solved.
Need to check novelty before this filing date? Find Prior Art