National Institute of Advanced Industrial Science and Technology (AIST)
Research resultsPublications > AIST TODAY > 2006-No.19
AIST TODAYNo.19 Winter 2006 [ PDF:17.6MB ]


A New Scheme of Adaptively Trainable Motion Image Recognition
- Achieved World's Best Performance for Human and Motion Recognition -

Nobuyuki Otsu
AIST Fellow
e-mail address

A new scheme has been developed for automatic recognition of persons and movements out of a monitored video image, a key step of automatic video-surveillance such as an anti-crime camera. The method is an extension of adaptive learning recognition scheme based on the Higher-order Local Auto-Correlation (HLAC) feature extraction developed for two-dimensional static images. To cover the feature extraction of "target's movements" in motion images, HLAC was extended to Cubic HLAC (CHLAC). The technology is characterized by enhanced versatility, high speed and high accuracy.

http://www.aist.go.jp/aist_e/latest_research/2005/20050621/20050621.html

Photo
Figure
Figure: Examples of human abnormal movement detection (Normal: walking through, Abnormal: tumbling over).

Relational Information

AIST TODAY Vol.5 , No.10 (2005) p.16-17



 back