Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection
We expect that gait will be useful information for detecting wandering and managing the health of elderly people utilizing a monitoring support robot. However, some kind of sensor is needed to extract pedestrian gait features. In this paper, we proposed a privacy-aware method of extracting gait feat...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | Japanese |
| Published: |
The Japan Society of Mechanical Engineers
2025-04-01
|
| Series: | Nihon Kikai Gakkai ronbunshu |
| Subjects: | |
| Online Access: | https://www.jstage.jst.go.jp/article/transjsme/91/945/91_25-00049/_pdf/-char/en |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849731845842796544 |
|---|---|
| author | Junya KOBAYASHI Nobuaki NAKAZAWA |
| author_facet | Junya KOBAYASHI Nobuaki NAKAZAWA |
| author_sort | Junya KOBAYASHI |
| collection | DOAJ |
| description | We expect that gait will be useful information for detecting wandering and managing the health of elderly people utilizing a monitoring support robot. However, some kind of sensor is needed to extract pedestrian gait features. In this paper, we proposed a privacy-aware method of extracting gait features from pedestrians’ feet only. This paper describes methods for detecting frames in which heel-strike and toe-off events occurred in foot video, and for extracting the heel contact position in the image. Firstly, dynamic regions were extracted utilizing edge detection and optical flow, and then clustering was used to extract pedestrians’ feet regions. Subsequently, the acceleration field was estimated using optical flow, the acceleration was decomposed into tangential and radial components, and the radial component acceleration was used to detect heel-strike frame. Next, we extracted static regions in the pedestrians’ feet region by utilizing edge detection, optical flow, clustered foot regions, and motion characteristics during gait, and finally performed toe-off frame detection. The heel contact position in the image was extracted by using Otsu’s binarization and static region. We conducted gait experiments with several camera direction conditions and applied it to the system. The RMSE of the estimated and true values for heel-strike and toe-off frame detection were within about one frame in all conditions, and the F-measure was above 80 % in all conditions. The F-measure for heel contact position detection also exceeded 80 % in all conditions. |
| format | Article |
| id | doaj-art-14db7b2255e24c6c9abb090fce8a8992 |
| institution | DOAJ |
| issn | 2187-9761 |
| language | Japanese |
| publishDate | 2025-04-01 |
| publisher | The Japan Society of Mechanical Engineers |
| record_format | Article |
| series | Nihon Kikai Gakkai ronbunshu |
| spelling | doaj-art-14db7b2255e24c6c9abb090fce8a89922025-08-20T03:08:25ZjpnThe Japan Society of Mechanical EngineersNihon Kikai Gakkai ronbunshu2187-97612025-04-019194525-0004925-0004910.1299/transjsme.25-00049transjsmePropose of heel-strike and toe-off frame detection from gait video with optical flow and edge detectionJunya KOBAYASHI0Nobuaki NAKAZAWA1Graduate School of Science and Technology, Gunma UniversityGraduate School of Science and Technology, Gunma UniversityWe expect that gait will be useful information for detecting wandering and managing the health of elderly people utilizing a monitoring support robot. However, some kind of sensor is needed to extract pedestrian gait features. In this paper, we proposed a privacy-aware method of extracting gait features from pedestrians’ feet only. This paper describes methods for detecting frames in which heel-strike and toe-off events occurred in foot video, and for extracting the heel contact position in the image. Firstly, dynamic regions were extracted utilizing edge detection and optical flow, and then clustering was used to extract pedestrians’ feet regions. Subsequently, the acceleration field was estimated using optical flow, the acceleration was decomposed into tangential and radial components, and the radial component acceleration was used to detect heel-strike frame. Next, we extracted static regions in the pedestrians’ feet region by utilizing edge detection, optical flow, clustered foot regions, and motion characteristics during gait, and finally performed toe-off frame detection. The heel contact position in the image was extracted by using Otsu’s binarization and static region. We conducted gait experiments with several camera direction conditions and applied it to the system. The RMSE of the estimated and true values for heel-strike and toe-off frame detection were within about one frame in all conditions, and the F-measure was above 80 % in all conditions. The F-measure for heel contact position detection also exceeded 80 % in all conditions.https://www.jstage.jst.go.jp/article/transjsme/91/945/91_25-00049/_pdf/-char/engait feature extractiongait event detectiongait analysisimage processingoptical flow |
| spellingShingle | Junya KOBAYASHI Nobuaki NAKAZAWA Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection Nihon Kikai Gakkai ronbunshu gait feature extraction gait event detection gait analysis image processing optical flow |
| title | Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection |
| title_full | Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection |
| title_fullStr | Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection |
| title_full_unstemmed | Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection |
| title_short | Propose of heel-strike and toe-off frame detection from gait video with optical flow and edge detection |
| title_sort | propose of heel strike and toe off frame detection from gait video with optical flow and edge detection |
| topic | gait feature extraction gait event detection gait analysis image processing optical flow |
| url | https://www.jstage.jst.go.jp/article/transjsme/91/945/91_25-00049/_pdf/-char/en |
| work_keys_str_mv | AT junyakobayashi proposeofheelstrikeandtoeoffframedetectionfromgaitvideowithopticalflowandedgedetection AT nobuakinakazawa proposeofheelstrikeandtoeoffframedetectionfromgaitvideowithopticalflowandedgedetection |