Abstract
Development of monitoring systems are critical for ensuring safety of operators in automated factories, since the number of such factories is expected to increase. However, methods for recognizing states of operators, such as postures and signals in case of emergency, have not yet been well studied. Here, the focus is on an image processing method for identifying operators by using several patterns and marks attached to their uniforms, recognition of postures and signals of operators. Several striped patterns and marks are chosen as plain identifiers of operators, and criteria of image processing are formulated for the identification of patterns and marks. In addition, two indices based on area and circumscribed rectangles of operators are introduced in order to judge their states, such as moving or standing, human or vehicle. Moreover, two ratios based on area of workers and their circumscribed rectangles are also applied to recognize operators' postures and signals in case of emergency. Workers' signals considered here are "stop", "need help", "OK", and "pay attention".
