Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

A very low complexity algorithm for detection a significant and rapid event in a geographically anchored and compressed video by using the length of the prediction frames and the corresponding telemetry

IP.com Disclosure Number: IPCOM000199650D
Publication Date: 2010-Sep-14
Document File: 1 page(s) / 22K

Publishing Venue

The IP.com Prior Art Database

Abstract

Extracting simple features (like frame rate etc) from real life video feed can be done with almost zero added processing. Merging it with telemetry meta-data (location, rotation, etc) enables us to detect significant events in a video from a moving camera and achieve very minimal amount of computation power

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 66% of the total text.

Page 1 of 1

Ȉ

Ȉ Ȉ

The problem is how to detect,

several kinds of significant events in a compressed video feed. Examples of such events are sudden flooding or fire works.

    Current methods requires at least parsing of the compressed video stream and most probably also processing of the decoded video frames.

Such processing usually make using these methods impractical.

    The complexity is very important when it is needed to detect significant events in many video feeds especially when the video compression is done in separate hardware, thus limiting the general purpose computing resources to a minimum.

Search words:

     geographically anchored video, telemetry , significant, event detection, rapid, motion detection, compressed-domain features, low complexity, p-frame length, prediction frame length, intra frame length, i-frame length, zoom in, zoom out, rotation, panning, rapid movement, height, pitch, yaw, roll, field of view

    Prior art: http://

www.ctr.columbia.edu/papers

www.springerlink.com/content/5835

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.89.8067&rep=rep1&type= pdf

    The core idea is to use the length of prediction frame / length of an average intra frame (or ant other readily available source) measure for detecting significant events in video periods where there isn't rapid zoom in/our, movement or rotation.

    The telemetry information will be used for detecting video periods with rapid zoom in/our, movement or rotation.

    Therefore we get a simple algorithm wit...