Method for analysis of object motion in videosTechnology #2017-059
Questions about this technology? Ask a Technology Manager
- Image Gallery
- Prof Chua Tat Seng NExT Centre for Extreme Search E-mail: email@example.comExternal Link (www.nextcenter.org)
- Managed By
- Ms Asha Srinivasan Associate Director (65)65161671
Novel method for analysis of object motion in videos
Object proposal, as a foundation of various multimedia applications, aims to generate a number of bounding boxes to indicate the object candidates in images and videos. From the generated proposals, object features can be extracted and analyzed for subsequent processing, such as object classification and information retrieval. However, selecting a video frame covering all the objects in a video is intractable, for the existence of camera motion and object motion can cause some objects omitted in the initial frame and thus excluded in the final proposals. Moreover, the strategy only uses the static appearance characteristics of video content and completely ignores its temporal motion characteristics, which are important for locating moving objects.
To address the aforementioned problems, a novel method is proposed for object trajectory proposal, which can effectively generate trajectory proposals for both moving and static objects by exploring the appearance and motion characteristics of video content.
Stage of Development
Analytical and laboratory studies to validate analytical predictions
All kinds of object-level video analysis, such as spatio-temporal characteristic extraction of objects and generic object detection. It can also be used to support the related applications, such as object-level video search, advertisement and recommendation.
Compared to existing solutions, our object
trajectory proposal can effectively handle both moving and static objects in videos
Our method significantly outperforms two state
of the art video object proposal methods..
Our method explore object motion to generate
trajectory proposals and is able to handle large object as well as comples
Patent pending. Available for licensing.
Prof Chua Tat Seng
NExT Centre for Extreme Search
ILO Ref No: 2017-059