Abstract
Agricultural workers can find it difficult to train new workers and to stay safe. To overcome these problems, we developed wearable sensors to record and analyze workers’ motions. To gather background information, we reviewed a wide range of sources (e.g., industrial goods, patents, and academic papers), discussed needs with agricultural informatics researchers, workers, and farm managers, and selected devices to monitor workers’ actions. First, we measured acceleration and angular velocity when subjects posed in typical postures and performed typical actions. Next, on small to medium-size farms, we logged the acceleration and angular velocity of the lower arms, chest, waist, and tools. A laptop computer in a knapsack on the back of each subject gathered and analyzed the data, using statistical and human dynamics software that we wrote. At the same time, we recorded the users’ actions from a few meters away on a fixed digital video camera. We then analyzed the data, mainly by optical flow methods. By analyzing various statistical differences concerning acceleration data between experienced and inexperienced workers and by showing (illustrating) the results to the workers for checking them and for utilizing them to revise workers’ movements, we confirmed the validity and usefulness of the system. The results, which need to be validated more precisely, will enhance workers’ skills and safety, and allow critical situations (e.g., risks of falling and injury) to be diagnosed. These techniques could be applied to other areas, being economical, robust, scalable, and easily distributed.
