Media.Player

Method and Apparatus for Detecting, Tracking, Recognizing, and Categorizing Objects and Scenes from Video

Technology Benefits

Feature selection and tracking are performed in real-time Improved baseline algorithms in terms of performance accuracy and robustness Lower computational costs

Technology Application

Interactive mobile applications including gaming, video editing, and image processing Security applications including image recognition and image data filtering

Detailed Technology Description

Researchers at UCLA have developed an end-to-end visual recognition system that allows a mobile user to select objects in their own real-time video for future visual recognition by the system. The method categorizes and tracks selected objects by selecting multiple feature points corresponding to translational, similarity, or affine reference frames, and processes the information with optimized algorithms. The system is then able to recall the selected objects in future real-time video streams.

Industry

ICT/Telecom

Sub Category

Multimedia

Application No.

8717437

Others

State Of Development

Researchers have developed a working prototype. To see the prototype in action, please follow this link on Youtube:
http://www.youtube.com/watch?v=cMv-McHw660&feature=player_embedded

Background

Almost all smart phones have built in video cameras for taking photographs and recording videos. Innovative applications have made it possible for users to enjoy the footage in many new ways. However, there are many more avenues to explore with built-in video cameras. In particular, current devices are not equipped to recognize, track, and process objects captured in a video stream. Such information is extremely useful for the future of gaming, security, video editing, and image processing.

Additional Technologies by these Inventors

• Dsp-Sift: Domain-Size Pooling For Image Descriptors For Image Matching And Other Applications
• Robust Visual-Inertial Sensor Fusion For Navigation, Localization, Mapping, And 3D Reconstruction

Tech ID/UC Case

22256/2011-479-0

Related Cases

2011-479-0

*Abstract

UCLA researchers in the Department of Computer Science have developed a real-time visual recognition system for mobile devices that allows a user to select objects in the video stream and the system tracks and categorizes the object for future recognition.

*IP Issue Date

May 6, 2014

*Principal Investigator

Name: Taehee Lee

Department:

Name: Stefano Soatto

Department:

Country/Region

USA