Depth dataset construction method, safety monitoring method, and related devices
By constructing parameterized simulation scenarios and sensor noise models, a deep dataset matching the real environment is generated, which solves the problem of insufficient monitoring accuracy in existing technologies and improves the training effect of safety monitoring models and the safety of equipment operation.
CN122154242APending Publication Date: 2026-06-05GUANGDONG LAB OF ARTIFICIAL INTELLIGENCE & DIGITAL ECONOMY (SZ) +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG LAB OF ARTIFICIAL INTELLIGENCE & DIGITAL ECONOMY (SZ)
- Filing Date
- 2026-04-30
- Publication Date
- 2026-06-05
Smart Images
Figure CN122154242A_ABST
Abstract
The application discloses a kind of depth data set construction method, safety monitoring method and related equipment, depth data set construction method includes: obtaining the scene configuration parameter corresponding to industrial safety scene;Based on scene configuration parameter, parameterized simulation scene is constructed;In response to preset safety event script, human body model and equipment model in simulation scene are controlled to interact, to generate corresponding safety event scene sequence;Based on sensor physical noise model, depth rendering is carried out to safety event scene sequence, and depth data is obtained;Based on scene geometric information and object attribute information, the label data corresponding to depth data is generated, to construct the depth data set corresponding to industrial safety scene.The depth data set of industrial safety scene constructed by utilizing sensor physical noise model for depth rendering, so that depth data can simulate real sensor measurement error, to improve the training effect and monitoring accuracy of industrial safety monitoring model.
Need to check novelty before this filing date? Find Prior Art