A computer vision-based motion time study for assembly process

Time and motion study is employed by manufacturing industries to determine work element time and identify operators’ ineffective motions. Accurate element time is crucial for effective process improvement and production planning. Operators’ ineffective motions lower their efficiency. By identifying and reducing or eliminating these motions, the operation time can be reduced and productivity can be increased. Traditional time and motion studies are conducted by human analysts using stopwatches, which may be subject to human errors. In this study, an automated time and motion study model is proposed. The model is comprised of computer vision and machine learning tools that work together to detect objects and events that identify the start and end of the work elements. This allows the computer to automatically time the work elements from a video. Once the time data are obtained, the model calculates the operator’s hand motion efficiency and summarizes the results to the users. A case study of a power extension socket assembly operation that implements the proposed model is presented. The time and motion efficiency results from the model are statistically compared to the time obtained by human analysts, as well as the reference values. Analysis result indicates that the model performs as well as human analysts in terms of element time and motion efficiency value accuracy, and is more reliable in terms of precision. Motion efficiency results also lead to operation improvement for the case study operation. To further demonstrate the benefit of using computer vision in the industries, a computer vision model system that can automatically perform material counting is also proposed.