A New Method for Continuous Posture Analysis Using Marker-based Video Processing Technique

Many postural analysis techniques are developed to reduce the risk of musculoskeletal problems recently. Methods such as Rapid Entire Body Assessment (REBA) are capable of analyzing the static or awkward positions, but the selection of these postures is subjective. To make an objective postural analysis, devices such as Electro Magnetic Trackers (EMTs) can be used continuously during the job task, but utilizing such devices is costly. Other methods such as Red, Green, Blue, and Depth (RGB-D) cameras and Microsoft Kinect have also been used in multiple studies to continuously track human posture. However, the mentioned methods are only capable of representing a single joint angle for body segments. Therefore, in this study a cost effective marker-based video processing algorithm is developed for measuring 3 Dimensional (3D) information regarding both the location and the orientation of human posture. The checkerboard pattern is selected as the markers to be detected in the video frame and relevant 3D location and orientation are calculated based on the Perspective n Point (PnP) approach. This study also investigates the precision of the measurements. Thus, an experiment was designed to capture markers in different known locations and orientations. Moreover, to validate the method of measurement, the second experiment is designed to compare the outcome of the proposed algorithm with the outcome of EMTs provided in another study. According to the result of the first experiment, overall, the algorithm was capable of measuring the 3D location and orientation of the markers with precision of 2.88 mm and 1.34° on average, respectively. Furthermore, the precision of the algorithm is found to be significantly affected by the marker pattern (p < 0.001). The higher the number of rows and columns showed a more precise measurement. The result achieved from the second experiment shows no significant difference between the measured value of the algorithm and EMTs (p = 0.880). Although some of the limitations of the proposed algorithm such as undetectable checkerboards at extreme angles relative to the camera, and limited field of view of the camera, promising results are achieved with a low cost of implementation.