On classification of acceleration and its components in computer vision for heel strike detection
On classification of acceleration and its components in computer vision for heel strike detection
In some forms of gait analysis, it is important to be able to localise the heel within the frame in which the strike occurs. According to the motion characteristics of heel strikes, radial acceleration is ideal for estimating the spatio-temporal position of heel strikes in standard image sequences. Previous research of motion analysis has generally not yet considered the basic nature of higher orders of motion such as acceleration. Hence, in this thesis, acceleration first is 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 motion in real world violate the global smoothness assumption. Further, we decompose acceleration into radial and tangential based on geometry.
Compared with previous heel strike detection techniques, acceleration not only improves the precision significantly but also enables detection in real-time. Our new method also shows a good robustness in performance analysis with respect to noised image and occlusion. Acceleration is propagated as a general motion descriptor, it shows the capability for differentiating different types of motion both on synthesised data and real image sequences.
Beyond acceleration, the higher-orders of motion flow and their continuant parts are preliminarily investigated for further revealing the chaotic motion fields. Naturally it is possible to extend this notion further: to detect higher orders of image motion. In this respect we show how jerk and snap can be obtained from image sequences. The derived results on test images and heel strike detection illustrate the ability of higher-order motion, which provide the basis for the following research and applications in the future.
University of Southampton
Sun, Yan
15039395-4052-4cb3-b2a8-1c7b08eff821
Sun, Yan
15039395-4052-4cb3-b2a8-1c7b08eff821
Hare, Jonathon
65ba2cda-eaaf-4767-a325-cd845504e5a9
Nixon, Mark
2b5b9804-5a81-462a-82e6-92ee5fa74e12
Sun, Yan
(2018)
On classification of acceleration and its components in computer vision for heel strike detection.
University of Southampton, Doctoral Thesis, 109pp.
Record type:
Thesis
(Doctoral)
Abstract
In some forms of gait analysis, it is important to be able to localise the heel within the frame in which the strike occurs. According to the motion characteristics of heel strikes, radial acceleration is ideal for estimating the spatio-temporal position of heel strikes in standard image sequences. Previous research of motion analysis has generally not yet considered the basic nature of higher orders of motion such as acceleration. Hence, in this thesis, acceleration first is 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 motion in real world violate the global smoothness assumption. Further, we decompose acceleration into radial and tangential based on geometry.
Compared with previous heel strike detection techniques, acceleration not only improves the precision significantly but also enables detection in real-time. Our new method also shows a good robustness in performance analysis with respect to noised image and occlusion. Acceleration is propagated as a general motion descriptor, it shows the capability for differentiating different types of motion both on synthesised data and real image sequences.
Beyond acceleration, the higher-orders of motion flow and their continuant parts are preliminarily investigated for further revealing the chaotic motion fields. Naturally it is possible to extend this notion further: to detect higher orders of image motion. In this respect we show how jerk and snap can be obtained from image sequences. The derived results on test images and heel strike detection illustrate the ability of higher-order motion, which provide the basis for the following research and applications in the future.
Text
Final Thesis
- Version of Record
Text
Permission to deposit unsigned
Restricted to Repository staff only
More information
Submitted date: November 2018
Identifiers
Local EPrints ID: 456053
URI: http://eprints.soton.ac.uk/id/eprint/456053
PURE UUID: 006cab06-c554-4fb1-926d-7a280bb8ad3d
Catalogue record
Date deposited: 25 Apr 2022 16:38
Last modified: 19 Sep 2024 01:39
Export record
Contributors
Author:
Yan Sun
Thesis advisor:
Jonathon Hare
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics