Active Vision System for Unusual Motion - PowerPoint PPT Presentation

1 / 11
About This Presentation
Title:

Active Vision System for Unusual Motion

Description:

... they did not study the detection of unusual motion and extraction of the ... After adding the time axis (30 frames), we get a cube for each sub-region. ... – PowerPoint PPT presentation

Number of Views:34
Avg rating:3.0/5.0
Slides: 12
Provided by: cspeo
Category:

less

Transcript and Presenter's Notes

Title: Active Vision System for Unusual Motion


1
Active Vision System for Unusual Motion
  • By Jing Zhong
  • CS580 Image and Video Computation
  • Spring 2002

2
Overview
  • Active Vision
  • An active vision system is one that is able to
    interact with its environment by altering its
    viewpoint rather than passively observing it, and
    by operating on sequences of images rather than
    on a single frame.
  • Three Types of Temporal Events
  • Temporal Textures, Activities, and Motion Events

3
Outline
  • Goal
  • In this project, we will use a monocular camera
    system to detect and track the area of unusual
    motion in the Field of View.
  • Background
  • The idea is inspired by the attentive mechanism
    of human vision. Previous work use camera control
    to realize active vision. However, they did not
    study the detection of unusual motion and
    extraction of the objects exhibiting unusual
    motion

4
Active Vision System
  • Attention
  • The attention mechanism in this project is driven
    by the unusual motion detection. We will build
    a simple statistical model for the unusual
    motion.
  • Fixation
  • Control the camera systems so that the
    attention is located at the center of the
    camera image plane and keeps the size proportion
    of interested object in the image invariably.
  • Pursuit
  • Keeping the fixation point as close as possible
    to the attention during the object or camera
    movement.

5
Current Work
y
  • Motion Flow Field Computation
  • Based on Intel OpenCV Library
  • Flow Field Decomposition
  • Decompose a bunch of frames into small
    rectangular sub-regions. After adding the time
    axis (30 frames), we get a cube for each
    sub-region.
  • Feature Extraction
  • Direction and Magnitude of Optical Flow
  • Generate empirical distribution in each cube
    (histogram)
  • Distance Computation

x
t
6
Video 1 and Motion Flow Field Map
7
Histogram(usual motion)
Histogram(unusual motion)
Unusual Motion Cube
Distance
8
The Histogram Distribution
185 Bin, 37 for the direction of the optical
flow, 5 for magnitude of the optical flow.
Usual Motion
Unusual Motion
9
Video 1 and Motion Flow Field Map
Failed
10
Unusual Motion Cube
11
Future Work
  • Using better distance measures
  • Using more robust of algorithm to build motion
    model
  • Combine tracking algorithm to track the unusual
    object
  • Dynamic camera control, so that we can zoom in
    and track the unusual motion
Write a Comment
User Comments (0)
About PowerShow.com