Industrial 3D Motion Capture Dataset Generation for Ergonomic Risk Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ergonomic risk assessment methods are labor-intensive and time-consuming, often failing to provide timely interventions for correcting workers' postures due to insufficient and unsuitable datasets for industrial settings, leading to inadequate 3D pose estimation performance of machine learning models.
Innovation Solution
A system and method using reflective markers, motion capture cameras, and visible light imaging sensors to generate comprehensive 3D motion capture datasets, including auto-labeling and gap filling, for calculating ergonomic angles in industrial contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If intermittent observation by ergonomic specialists is used, then expert assessment can be obtained, but the method is labor-intensive and time-consuming, preventing timely intervention
Solution Approach 1:
The system enables automatic ergonomic risk assessment through computer vision and machine learning models that continuously analyze worker postures without requiring expert intervention. The automated pose estimation and risk calculation allow the system to self-evaluate ergonomic conditions in real-time, eliminating the need for manual specialist assessment while maintaining assessment quality.
Solution Approach 2:
The patent replaces the mechanical system of manual observation and measurement with an automated computer vision system using cameras, deep learning models, and algorithmic pose estimation. This substitution transforms ergonomic assessment from a labor-intensive manual process to an automated digital system that operates continuously without human intervention.
2Ease of manufacture
If simplified skeleton representations are used in datasets, then data processing is easier, but essential keypoints information for calculating intricate ergonomic angles is lost
Solution Approach 1:
The patent implements a detailed skeleton representation that captures local quality and specificity of body keypoints, particularly focusing on wrist, elbow, shoulder, and spinal joints. Rather than using simplified generic skeletons, the system identifies and tracks specific anatomical landmarks with high precision to enable calculation of intricate ergonomic angles while maintaining comprehensive information about body posture.
Solution Approach 2:
The system transitions from 2D image data to 3D pose estimation by incorporating depth information and spatial relationships. This dimensional enhancement allows the model to calculate three-dimensional ergonomic angles and postures, providing comprehensive keypoints information that simplified 2D representations cannot capture, while still maintaining computational feasibility through efficient 3D reconstruction algorithms.
3Ease of manufacture
If generic motion datasets are used for training, then model development is simpler, but the models perform poorly when deployed in industrial settings
Solution Approach 1:
The patent collects and annotates motion capture data specifically from industrial workers performing actual work tasks before deploying the model. This preliminary data collection phase involves capturing 3D motion data, labeling keypoints, and creating a specialized training dataset that reflects real industrial conditions. By preparing domain-specific training data in advance, the model learns the specific variations and characteristics of industrial postures, ensuring reliable performance when deployed in the target environment.
Solution Approach 2:
The system adapts the training process by changing key parameters including data collection methodology, annotation protocols, and model architecture to suit industrial settings. The patent modifies the dataset composition to include diverse industrial tasks, adjusts pose estimation parameters for industrial lighting and camera conditions, and tunes model hyperparameters based on industrial worker characteristics, thereby optimizing model performance for the specific deployment environment.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Provides a detailed and accurate dataset for ergonomic risk assessment, enabling continuous and timely identification of awkward postures, improving the performance of machine learning models in industrial environments.
Implementation Method 1
a set of motion capture cameras placed in first selected locations of an area to detect signals from the plurality of reflective markers
Data Source
AI summary
Disclosed herein is a system and method for obtaining and generating motion capture datasets relating to various working activities for ergonomic risk assessment. An example system may comprise a computing device that obtains first data from multiple motion capture cameras, obtains second data from multiple visible light imaging sensors, calculates 3D positions of multiple reflective markers positioned on several subjects performing various working activities based on the first data, labels each marker to generate marker trajectories, performs gap filing and smoothing functions on the marker trajectories to generate global marker positions, transforms the global marker positions into a corresponding image coordinate system of each sensor to generate 3D pose data of the subjects at each sensor viewpoint, projects the 3D pose data into frames of the second data to generate 2D pose data, and generates a dataset comprising the second data, the 2D pose data, and the 3D pose data.