As there has been an increasing interest in work-related musculoskeletal disorders (MSDs) due to repetitive motion in various industries, many efforts have been made to prevent these MSDs in view of ergonomics. The first step toward the prevention of the MSDs is the accurate measurement of working postures which is followed by their analysis that can be done by one of methods such as NIOSH Lifting Equation (NLE), Ovako Working Posture Analysis System (OWAS), Rapid Upper Limb Assessment (RULA), Rapid Entire Body Assessment (REBA), etc.
There are largely two types of methods for measuring and analyzing working postures: image-based and motion-based. The image-based method using the captured images of working postures is relatively inexpensive. However, it is operated by the analyst with knowledge in ergonomics who usually needs much time and effort in image recording and analysis. Also, the analyst’s subjective opinions or mistakes may be reflected in the results, which are relatively less reliable. The motion-based method can produce accurate and consistent results since it uses a motion capturing and analysis system, which precisely measures work postures and analyzes them based on mathematical computation. However, the system is not always affordable and available as it is very expensive and requires a large space for its use.
Since the recent introduction of Kinect, which is a low-cost device for detecting and measuring the user’s motion, there have been various research works using this. In this thesis, as one of efforts to prevent MSDs, we propose a method for analyzing working postures using Kinect and Augmented Reality (AR) markers. Using the Kinect, we measure a worker’s posture and obtain its corresponding skeleton model. We use the captured skeleton model not only to animate a 3D human model in Unity3D but also to compute a set of values required to apply the REBA that assesses the workload of a given working posture. Moreover, we properly combine an AR marker tracking technique with the use of the Kinect in order to determine which side of the worker’s body the Kinect is facing and to measure the turning angle of the worker’s neck, which are not detectable only by the use of the Kinect. To show the quality and usefulness of the proposed method, we have implemented it and conducted experiments including its comparison to the image-based method in actual cases. As a result, we have found that the proposed method is fairly cost-efficient, decently accurate, and able to run in real-time.