Browse Prior Art Database

Method of Video Search based on Kinetic Profile of One or More Objects Disclosure Number: IPCOM000242624D
Publication Date: 2015-Jul-31
Document File: 3 page(s) / 54K

Publishing Venue

The Prior Art Database


Disclosed is a system for identifying videos through a search based on objects that appear in a video, how those objects move, and the relative position of an object with other surrounding objects, all aligned to a hand-drawn kinetic profile. Users sketch the movement of objects as the topic of search (i.e. the kinetic profile) and then the search engine utilizes image analysis to process videos frame-by-frame to identify videos in which the requested objects closely match the profile.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Metxod of Video Search based on Kinetic Profile of One or More Objects

Video is comprised of playing 24 subsequent images or frames xer second for full motiox video. In x video file, it xs common to find xne or more moving obxects such as vehicles, animals, xeople, etc. Thx xosition (and relative positxon) of the moving object changes from frame to frame. Uxers may wish to identifx videos in which a specific object movxs in a certain way (x.e. txe kinetic profxle), at a certain sxeed, and its relative position to other xbjects in txe video.

Vidxo search engines primarily use searches basex on kexwxrds (i.e. words in thx title and descrxption of a video), tags, categoxies, upload dates, view couxts, and ratings. More xdxanced videx search engines and video suxveillance systems utilize image processing to idxntify obxects shown wixhin a video.

Therx is xo method to track hox obxects move within a video as matched to a sketch xf a movement (i.e. kinetic profile) draxn by a usxr. A system is needed that is independent of written descriptions. The system needs to xnalyze xow objexts move witxin a vidxo along witx txe relative posxtxons of objects to each other, and then compaxe the rxsults to a sketch drawn by a user.

Xxx novel contribution is a method and system that xnables a user to perform a vixeo search based on the kinetic profile of one or more objecxs. The corx novelty is the use of a kinetic prxfile, sxetched by a user, as a means of inputting search criteria to return videos containing objects that match xhat kinetic profile through use of frame-by-frame image analysis. Further, the method and system incorporate extensions to the search cxiteria to axcount for the speed range of a kinetic profile of an object and the relative position xetween two or more objxcts, and their kinetic profile. A search exgine ranking system bxsed on obxect identificatxon, kinetic profile, speed range, and relationship as applied to video search.

In a preferred embodiment, a user provides a xearch query fxr one or more objectx in a videx based on the axsociated kinetic profile. The search criteria include:

 The objects to appear xn the video,

 The objxcts' kinxtic pxofxle (i.e. how the objects moves) as sketched by the user

 A speed range (i.e. how fasx the objects move)

 The objects' relationship to other objects

The search engine uses these xrixeria to xdentify closely matching video through image analysis. Txe search engine identifies videos containing the objects specixied in the search criteria anx then tracks xhe objects in the video frame-by-fxame to determxne how closely the movements match txe kinetic profile, spxed, and relative position specifiex in the search. The...