Computer vision based extraction of experimental parameter mediated by time series analysis

Abstract Data retrieval and feature extraction by means of different computational techniques has played a vital role in the field of artificial intelligence as it has been successfully employed for the extraction of useful parameters from the input data. However, successful implementation of a relatively simplified approach for the experimental parameter extraction from the visual data is yet to be explored. In this work, we implement a computer vision (CV)-based automatic video analysis technique for the estimation of time period of a simple pendulum from the visual data of the performed experiment. We adopt various direct and indirect methods including real-time measurement of oscillation time, visual estimation through recorded video clips, and through CV-generated time series graphs of oscillations. The analysis of the time series plots by using fast Fourier transform helps in determining the frequency and consequently the time period of the harmonic oscillations. The comparison of the resulting values estimated by all of the implemented methodologies with the one theoretically calculated, shows that the CV based technique performs equally well with a reasonable level of accuracy and tops the list of adopted methods in terms of the least human intervention. These findings may help in paving the avenue for future implementation of the CV-based strategies in parameter extraction from the visual data of experiments in the field of natural sciences.