Measurement of Seating Posture by Automatic Numeric Marker Detection using the Stereo Camera

Abstract

A proper seating posture on a wheelchair is important in order to avoid swallowing and respiratory problems. In addition, an improper seating posture may cause a pressure sore and affect motor performance. Although systems that can measure angles of the seating posture that are defined by the international standard ISO16840-1 have been introduced, there are many improvements that can be made in this regard. The system developed in this study aims to measure the seating posture in three dimensions by recognizing numeric markers pasted on the subject from the images captured by stereo video cameras. A system that can recognize the markers and measure the three-dimensional co-ordinates of the markers and angles of nodal lines composed by the markers was developed using MATLAB (MathWorks). By defining the relation between the up, down, left, and right numeric markers, and utilizing this information, the recognition ratio was improved. The measured co-ordinates and angles were compared with the data obtained by MAC3D (Motion Analysis Corporation). The maximum average co-ordinate errors reduced when a large 500 × 750 mm calibration board was used. When a skeletal model was used, the maximum average error in the X direction was 4.8 mm and the angle error was 0.5°. When five human subjects were used, the maximum average error in the X direction was 10.4 mm and the angle error was 1.6°. The errors increased because there was an interval of around 30 s from the measurement made by the developed system to the measurement by the MAC3D system, during which the subjects' body moved.