관절 데이터 기반 사람 동작 분류 모델 및 자세 위험성 평가 시스템

초록

The modern manufacturing industry faces various challenges of reducing labor costs, increasing productivity, and securing worker safety. For decades, much effort has been taken to solve these problems by introducing automation of the manufacturing process. Technical and economic limitations induced by fully automated processes employing industrial robots have brought the need of introducing collaborative robots. In order for robots and human operators to collaborate in manufacturing systems, it is desirable for the robot to be able to recognize human motions and to predict the subsequent motions. In spite of advances in classification techniques, the approaches based on motion data collected from the front and the side have suffered from relatively low consistency in classifying similar motions. In this study, we propose a model using angular transformation that can improve accuracy in recognition and classification of human joint data of frontal and lateral motions collected by motion capture equipment. Furthermore, a risk assessment of operators posture is presented based on the transformed human joint data.

키워드