Development of a measurement and evaluation system for bed-making activity for self-training

This study proposes a method to automatically measure multiple objects by image processing for constructing a system for nursing trainees of self-training in the skill of bed making. In a previous study, we constructed a system to measure and evaluate trainee performance using three RGB-D (RGB color and depth) sensors. Our previous system had a problem with recognition of equipment such as the bed pad and the sheet because of color change by the light condition, the automatic color correction by the sensors and color variability in one object. In this paper, we used color reduction and cluster selection for equipment recognition. The system reduced the color in images by using k-means clustering and recognized the clusters as separate objects by predetermined thresholds. Compared with the previous method, the recognition accuracy was higher and the accuracy achieved was 70%.