Development of Worker Position and Posture Detection System Using PTZ Camera for Intelligent Agricultural Work Assistance

Abstract

An intelligent agricultural worker assistance system that can track worker positions and behavior patterns and comprehend workflow presents a potential method of facilitating workload reduction and improving work efficiency. In this study, we developed an image processing system capable of simultaneously detecting both the position and posture of a worker on a route by combining a single pan-tilt-zoom camera, an original working uniform marked with four different red-blue combinative colors and two types of particle filters that permit detection of the color markers on the working uniform. In our experiments, we determined that the proposed system could detect the position of a worker that was walking, turning and moving his arms from standing and sitting postures on a monitored route in a field with a resolution of approximately 1 m and a sensing range of approximately 38 m. The system could also detect, with no exceptions, eight separate worker postures. The results of these experiments indicate that our developed system could be used for tracking worker positions and behavior patterns in real time.