Using image recognition to automate video analysis of physical processes

Abstract

[EN] Multimedia technologies and video analysis allow the design of low-cost physical measurement and data acquisition systems. The number of pixels in the images limits their spatial resolution, whereas their time resolution depends on the number of frames recorded per second. Both characteristics are determined by the video recording system used. We have developed a laboratory system with these characteristics where important improvements have been reached by using image recognition to automate video analysis. In the present work we first examine several image recognition techniques and evaluate them from the point of view of their application to measurement systems. Then we describe the proposed system and the methodology followed in the measurement process. An image of an experimental object is recorded and used as a filter, while the sequence of images of that object in motion become the input frames for the recognition process. Finally, we discuss the results obtained by this measurement process and compare them with those obtained by using traditional measurement techniques.