Measuring exertion time, duty cycle and hand activity level for industrial tasks using computer vision.

Area of Science:

• Occupational Health and Safety
• Computer Vision
• Ergonomics

Background:

• Manual assessment of industrial tasks is time-consuming and prone to human error.
• Accurate estimation of exertion time, duty cycle (DC), and hand activity level (HAL) is crucial for ergonomic evaluations.
• Existing methods for data collection can be inefficient and costly.

Purpose of the Study:

• To develop and validate computer vision algorithms for automatic estimation of exertion time, duty cycle (DC), and hand activity level (HAL).
• To compare the accuracy of computer vision algorithms against manual frame-by-frame analysis.
• To assess the impact of duty cycle estimation errors on hand activity level calculations.

Main Methods:

• Development of two computer vision algorithms: Decision Tree (DT) and Feature Vector Training (FVT).
• Application of algorithms to videos of workers performing 50 diverse industrial tasks.
• Comparison of algorithm outputs (DC, HAL) with manual frame-by-frame analysis.
• Conducting a sensitivity analysis to evaluate the effect of DC deviations on HAL.

Main Results:

• The Feature Vector Training (FVT) algorithm demonstrated a low average duty cycle (DC) difference of 1.4% compared to manual analysis.
• Both Decision Tree (DT) and FVT algorithms showed minimal average hand activity level (HAL) differences (0.5 and 0.3, respectively).
• Sensitivity analysis indicated that duty cycle errors below 5% had a negligible impact on hand activity level estimations.

Conclusions:

• Computer vision algorithms provide a reliable and automated method for estimating key ergonomic parameters like duty cycle and hand activity level.
• Automated estimation using computer vision is comparable in accuracy to traditional manual frame-by-frame analysis.
• The developed algorithms can enhance the efficiency and objectivity of ergonomic assessments in industrial settings.