Evaluating Reliability of Motion Features in Surveillance Videos

1
Evaluating Reliability of Motion Features in
Surveillance Videos

• Longin Jan Latecki and Roland Miezianko, Temple
  University
• Dragoljub Pokrajac, Delaware State University

November 2004
2
Motion Detection

• Goals of motion detection
• Identify moving objects
• Detection of unusual activity patterns
• Computing trajectories of moving objects
• Benefits of reliability assessment
• Reduction of false detections (e.g., false alarms)

3
Applications of Motion Detection

• Many intelligent video analysis systems are based
  on motion detection. Such systems can be used in
• Homeland security
• Real time crime detection
• Traffic monitoring

4
Motion Measure Computation

• We use spatial-temporal blocks to represent
  videos
• Each block consists of NBLOCK x NBLOCK pixels
  from 3 consecutive frames
• Those pixel values are reduced to K principal
  components using PCA (Kahrunen-Loeve trans.)
• In our application, NBLOCK8, K10
• Thus, we project 192 gray level values to a
  texture vector with 10 PCA components

5
Frame t-1
6
Frame t
7
Frame t1
8
443 spatial-temporal block Location I7, J7,
time t
48-component block vector (443)
9
Why texture of spatiotemporal blocks can work
better?

• More robust in comparison to pixel-based approach
• Integrates texture- and movement (temporal)
  information
• Faster

10
499
624
863
1477
11
Trajectory of block (24,8) (Campus 1 video)
Moving blocks corresponds to regions of high
local variance
Space of spatiotemporal block vectors
12
Trajectory of a pixel from block (24,8)
Space of RGB pixel values
13
Detection of Moving Objects Based on Local
Variation

• For each location (x,y) of the frames
• Consider vectors of derived attribute values
  corresponding to a symmetric window of size 2W1
  around each time instant t
• Derived attribute vectors RGB first 10 PCA
  projections of spatial-temporal blocks, etc.
• Compute the covariance matrix for the vectors
• motion measure is defined as the largest
  eigenvalue of the covariance matrix

14
Feature Vectors in Space
Feature vectors
4.2000 3.5000 2.6000 4.1000
3.7000 2.8000 3.9000 3.9000 2.9000
4.0000 4.0000 3.0000 4.1000 3.9000
2.8000 4.2000 3.8000 2.7000
4.3000 3.7000 2.6500
Covariance matrix
Current time
0.0089 -0.0120 -0.0096 -0.0120
0.0299 0.0201 -0.0096 0.0201 0.0157
Motion Measure
Eigenvalues
0.0499 0.0035 0.0011
0.0499
15
Feature Vectors in Space
Feature vectors
4.3000 3.7000 2.6500 4.4191
3.5944 2.4329 4.1798 3.8415 2.6441
4.2980 3.6195 2.5489 4.2843 3.7529
2.7114 4.1396 3.7219 2.7008
4.3257 3.6078 2.8192
Covariance matrix
0.0087 -0.0063 -0.0051 -0.0063
0.0081 0.0031 -0.0051 0.0031 0.0154
Current time
Motion Measure
Eigenvalues
0.0209 0.0093 0.0020
0.0209
16
In our system we divide video plane in disjoint
blocks (8x8 blocks), and compute motion measure
for each block.
mm(x,y,t) for a given block location (x,y) is a
function of t
17
Graph of motion measure mm(24,8,) for Campus 1
video
18
Motion amount
The feature called motion amount is defined as

• The system decision on alarm situation is based
  on ma.

19
(No Transcript)
20
ma(t) as function of frame number t for Temple 1
video
21
(No Transcript)