Research Achievements Of Construction Ergonomics

1721 Words Dec 21st, 2014 7 Pages
Research Achievements in Construction Ergonomics
As the rapid development and extensive applications of motion capture technology in the past few years, advances in WMSD assessment have been made in the construction research field, especially on the posture-oriented assessment. Among them, vision-based studies and wearable sensor systems are two main ongoing streams and the details are discussed respectively. Table 4 provides a state-of-the-art overview in achievements of a variety of works in construction in terms of sensors, focused risks, and data processing.
<Insert Table 4 here>
Vision-Based Methods
The vision-based method is a marker-less assessment method. It relies on depth sensors or multiple cameras to achievement 3D
…show more content…
Kinect outperforms the traditional optical-marker-based systems (e.g., Vicon) in its cost-effectiveness and high precision. Performance of Kinect in the construction safety monitoring was tested in a ladder climbing experiment of 25 trials of ascending and descending taken by a subject, and the error of motion data falls in an acceptable range compared to the ground truth provided by the Vicon system (Han et al. 2013). The Vicon data is regarded as ground truth. Through correspondence and synchronization, the Kinect data and Vicon data are mapped into the same coordinate system and a discrepancy of up to 10.7 cm in joint location and 16.7 degrees in rotation angle is detected. The performance of Kinect data for motion analysis and action recognition (i.e., ascending and descending) is also evaluated, and only 1 case of an ascending action was detected incorrectly (i.e., accuracies of 98% and 100% for ascending and descending actions, respectively). Kinect has proven the potential in daily risk assessment on construction sites (Han et al. 2012). However, range sensors have the downside of short range applicability (i.e., 0.8-4m). Besides, the classification and recognition are limited in simple gestures and motions in the restricted range. In addition, occlusion by objects and the high illumination can severely reduce the accuracy of classification. Therefore, future work is required to build an automated vision-based
Open Document