Sampling methods with least information loss in transit videos for the reduction of manual work and computational processing

Abstract

Automated object recognition in traffic videos is a complex and time-consuming task that requires not only computational processes, but also some manual labor. The amount of time spent on both processes and labor is closely related to the number of frames to be processed. In this research, various sampling methods were studied to reduce the number of frames. Systematic sampling of 29% of the frames is the one that uses the least number of frames and is also equivalent to the census in terms of recognition error.