lect1027-1029h - Change Detection Main Points Detect pixels...

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon
1 Change Detection Main Points • Detect pixels which are changing due to motion of objects. • Not necessarily measure motion (optical flow), only detect motion. • A set of connected pixels which are changing may correspond to moving object.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
2 Picture Difference = otherwise T y x DP if y x D i K K 0 ) , ( 1 ) , ( | ) , ( ) , ( | ) , ( 1 j y i x f j y i x f y x DP i i m m i m m j + + - + + = - - = - = ∑ ∑ | ) , ( ) , ( | ) , ( j y i x f j y i x f y x DP k i i m m i m m i m m k + + - + + = + - = - = - = ∑ ∑ ∑ | ) , ( ) , ( | ) , ( 1 y x f y x f y x DP i i - - = Background Image • The first image of a sequence without any moving objects, is background image. • Median filter )) , ( , ), , ( ( ) , ( 1 y x f y x f median y x B n K =
Background image of page 2
3 PFINDER Pentland Pfinder • Segment a human from an arbitrary complex background. • It only works for single person situations. • All approaches based on background modeling work only for fixed cameras .
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
4 Algorithm • Learn background model by watching 30 second video • Detect moving object by measuring deviations from background model • Segment moving blob into smaller blobs by minimizing covariance of a blob • Predict position of a blob in the next frame using Kalman filter • Assign each pixel in the new frame to a class with max likelihood. • Update background and blob statistics Learning Background Image • Each pixel in the background has associated mean color value and a covariance matrix. • The color distribution for each pixel is described by Gaussian. • YUV color space is used.
Background image of page 4
5 Detecting Moving Objects • After background model has been learned, Pfinder watches for large deviations from the model. • Deviations are measured in terms of Mahalanobis distance in color. • If the distance is sufficient then the process of building a blob model is started. Detecting Moving Objects • For each of k blob in the image, log- likelihood is computed ) 2 ln( 5 . | | ln 5 . ) ( ) ( 5 . 1 D m K y K y d k k k T k k - - - - - = - m m Log likelihood values are used to classify pixels )) , ( ( max arg ) , ( y x d y x s k k =
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
6 Updating •The statistical model for the background is updated. y y y E K t t T t t t a m a m m m + - = - - = - 1 ) 1 ( ] ) )( [( • The statistics of each blob (mean and covariance) are re-computed. Mixture of Gaussians Grimson
Background image of page 6
7 Algorithm • Learn background model by watching 30 second video • Detect moving object by measuring deviations from background model, and applying connected component to foreground pixels. • Predict
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 8
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 35

lect1027-1029h - Change Detection Main Points Detect pixels...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online