A Computer Vision-Based Model for Automatic Motion Time Study

Motion time study is employed by manufacturing industries to determine operation time. An accurate estimate of operation time is crucial for effective process improvement and production planning. Traditional motion time study is conducted by human analysts with stopwatches, which may be exposed to human errors. In this paper, an automated time study model based on computer vision is proposed. The model integrates a convolutional neural network, which analyzes a video of a manual operation to classify work elements in each video frame, with a time study model that automatically estimates the work element times. An experiment is conducted using a grayscale video and a color video of a manual assembly operation. The work element times from the model are statistically compared to the reference work element time values. The result shows no statistical difference among the time data, which clearly demonstrates the effectiveness of the proposed model.