Cloud based video detection and tracking system

Inventors

BLASCH, ERIK P.Liu, KuiLIU, BINGWEIShen, DanChen, Genshe

Assignees

United States Department of the Air Force

Publication Number

US-9373174-B2

Publication Date

2016-06-21

Expiration Date

2034-10-21

Interested in licensing this patent?

MTEC can help explore whether this patent might be available for licensing for your application.


Abstract

A method for detecting and tracking multiple moving targets from airborne video within the framework of a cloud computing infrastructure. The invention simultaneously utilizes information from an optical flow generator and an active-learning histogram matcher in a complimentary manner so as to rule out erroneous data that may otherwise, separately, yield false target information. The invention utilizes user-based voice-to-text color feature description for track matching with hue features from image pixels.

Core Innovation

The invention provides a method for detecting and tracking multiple moving targets from airborne video within a cloud computing infrastructure. It integrates information from an optical flow generator and an active-learning histogram matcher in a complementary manner to eliminate erroneous data that could otherwise produce false target information. Additionally, user-based voice-to-text color feature description is employed for track matching with hue features from image pixels.

The invention addresses the challenge of detecting and tracking moving objects in noisy outdoor video environments, such as those degraded by atmospheric turbulence and sensor platform scintillation. These conditions, combined with the fact that moving objects may occupy only a few pixels, make motion detection very difficult and prone to many false alarms using existing approaches.

To overcome these issues, the invention features a developed target detection and tracking system comprising cloud-based image alignment, online target color calibration, optical flow detection, and histogram matching. The system uses template hue histograms in HSV color space to represent target colors, actively calibrating these histograms during operation to adapt to varying lighting and background conditions. The cloud infrastructure enables parallel processing of computationally intensive tasks such as image registration, thereby achieving real-time frame rates with non-real-time algorithms.

Claims Coverage

The patent includes one independent claim describing a comprehensive method for video detection and tracking of a target, comprising several key steps and features.

Method for defining and representing a target color

Defining a target by selecting a dataset of target image frames and a desired color, and converting the color into a template hue histogram representation.

Image alignment and registration with homography matrix generation

Performing target image frame alignment and registration, including generating homography matrices to enable coordinate transformations across frames.

Optical flow field generation and morphology processing to identify candidate target contours

Generating an optical flow field between aligned consecutive frames and applying morphology processing to extract candidate target contours.

Matching candidate contours to template hue histograms for target recognition

Matching the candidate contours derived from optical flow data to the template hue histogram representation to identify potential targets.

Initialization and tracking of the target with sequential track formation

Initializing tracking and generating target tracks by aligning current target image frames and projecting prior tracks using rotation and translation matrices derived from homography matrices.

Online tuning and calibration of template hue histograms

Adapting color processing to light source characteristics and updating the template hue histogram in HSV color space through weighted linear addition with candidate histograms to improve robustness.

Use of cloud computing environment and web GUI with voice command interface

Performing computationally intensive steps such as image alignment within a cloud computing environment accessible through a graphical user interface and voice commands, enabling parallel processing and elastic resource allocation.

The independent claim covers a method combining target definition by color histogram, image registration, optical flow-based candidate detection, histogram matching for target recognition, tracking initialization and update with homography-based projection, and adaptive histogram calibration within a cloud computing framework supporting web and voice interfaces.

Stated Advantages

Increased robustness of target detection and tracking by combining optical flow generation and active-learning histogram matching to eliminate false alarms.

Adaptation to varying and difficult lighting and background conditions through online target color calibration using histogram tuning.

Real-time performance achieved by leveraging cloud computing for parallel processing of computationally intensive tasks.

User interaction enhanced via voice-to-text color feature description and a web graphical user interface for flexible and intuitive operation.

Documented Applications

Detection and tracking of multiple moving targets from airborne video for suspicious behavior recognition in defense and security applications.

Robust motion detection and target tracking in outdoor video environments affected by noise sources such as atmospheric turbulence and sensor platform instability.

JOIN OUR MAILING LIST

Stay Connected with MTEC

Keep up with active and upcoming solicitations, MTEC news and other valuable information.