Underground human body posture estimation model training method and device

CN119785412APending Publication Date: 2025-04-08HEFEI UNIV OF TECH
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202311302521.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-10-08
Publication Date
2025-04-08

AI Technical Summary

Technical Problem

The prior art is difficult to accurately identify human posture in mine underground environments with low light intensity and uneven light distribution, especially in time series data processing problems such as time delay and insufficient computing power.

Method used

By obtaining the depth map of the human body posture, inputting the initial feature extraction network after preprocessing, extracting the joint key map and standard key map, and combining the similar relationship score to train the action classification network to generate a human body posture estimation model. The model uses a depth camera to acquire depth information, combined with feature extraction and action classification networks, and can accurately identify human postures in low-light environments.

Benefits of technology

It realizes efficient and accurate identification of the postures of underground workers in low-light environments, improves detection accuracy, and can distinguish different postures with similar actions, which helps to standardize the behavior of underground personnel, prevent safety accidents, and ensure the normal operation of equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119785412A_ABST
    Figure CN119785412A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of human body posture estimation, in particular to an underground human body posture estimation model training method and device. Inputting the preprocessed human body posture depth map and the standard posture map into an initial feature extraction network, and respectively outputting a joint key map and a standard key map; generating a trained target feature extraction network by using the joint key graph and the standard action key graph; similarity relation scores of the joint key images and standard key images are calculated, a trained target action classification network is obtained, and the target feature extraction network and the target action classification network form the human body posture estimation model. According to the method, the detection precision of close-range personnel is improved, different postures with similar actions can be effectively distinguished, behaviors of underground personnel can be standardized, safety accidents are prevented, and normal operation of underground equipment is guaranteed.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • Train driver specific motion detection method

    CN108446586A

  • Multi-person evaluation system based on machine vision skeletal tracking

    CN115482580A

  • Human body posture image intelligent identification method and system

    CN115527269A