Analysing Acceleration for Motion Analysis

Abstract

Previous research in motion analysis of image sequences has generally not considered the basic nature of higher orders of motion such as acceleration. In this work, we disambiguate different types of motion, and in particular focus on acceleration. First, we show acceleration can be computed in a principled manner by extending Horn and Schunck's algorithm for global optical flow estimation. We then demonstrate an approximation of the acceleration field using an alternative established optical flow technique, since most real motions violate the global smoothness assumption of Horn and Schunck. Furthermore, we decompose acceleration into radial and tangential components for greater depth of understanding of the motion. As a general motion descriptor, we show how acceleration provides the capability for differentiating different types of motion in video sequences.

On parameterizing higher-order motion for behaviour recognition

“…Presenting acceleration, jerk and snap using their radial and tangential components can help understanding. Acceleration flow was decomposed in our previous work [5]. Further, jerk and snap flow are now resolved as well: the definition of tangential and radial components of jerk are the linear changes of tangential and radial components of acceleration, they are computed in a manner similar to Equation ( 5) and ( 6):…”

“…As the frame sampling rate in video sequence is fixed the motion fields between neighbouring frames is usually considered as velocity. Acceleration flow was derived from the basis of Horn-Schunk in our earlier work [5] and now it has been extended into the detection of jerk and snap (and in their vector format). The new algorithms are investigated both on synthetic and real images, providing a new angle to understand and disambiguate dynamic relationships in image sequences.…”

“…More recently, Edison and Jiji have proposde two optical acceleration algorithms and utilize them to recognize action but do not extend beyond acceleration [58]. In this paper we extend higher-order motion analysis and parameterizes different types of motion for behaviour recognition based our previous work [5], [6]. The new contributions of this paper are:…”