Title: Color-Invariant Motion Detection under Fast Illumination Changes
1Color-Invariant Motion Detection under Fast
Illumination Changes
- Paper byMing Xu and Tim Ellis
- CIS 750
- Presented by Xiangdong Wen
- Advisor Prof. Latecki
2Agenda
- gtIntroduction
- gtColor Fundamentals
- gtColor-Invariant Motion Detection
- gtExperimental Results
- gtDiscussion
- gtConclusion
3Introduction
- gtMotion detection algorithms based on the
differencing operation of image intensities
between a frame and a background image. - gtBackground image reflects the static elements
in a scene. - gtBackground image needs to be updated because
- gtLack of a target-free training period
- gtGradual illumination variations
- gtBackground objects which then move.
- gtUpdating scheme
- Linear interpolation between the previous bg
value and new observation - gtGaussian mixture model based on gray-level or
RGB color intensities - gtcould detect a large proportion of changes.
- gtcannot follow fast illumination changes
- gtmoving clouds,
- gtlong shadows,
- gtswitching of artificial lighting
4Identifying a particular object surface under
varying illumination
- gtMarchant and Onyango.(2000) proposed a
physics-based method for shadow compensation in
scenes illuminated by daylight. - gtRepresented the daylight as a black body,
- gtAssumed the color filters to be of
infinitely narrow bandwidth. - Results as illumination changes, the ratio
(R/B)/(G/B)A depends on surface reflection only.
(A can be calculated from daylight model and
camera.) - gtFinlayson et al.(2000) Using same scheme.
- Results Log-Chromaticity Differences (LCDs)
ln(R/G) and ln(B/g) are independent of light
intensity and there exists a weighted combination
of LCDs which is independent of light intensity
and light color
5Adaptive schemes in color-invariant detection of
motion under varying illumination
- gtWren et al (1997) used the normalised
components, U/Y and V/Y of YUV color space to
remove shadows in a indoor scene - gtA single adaptive Gaussian represents the
probability density of pixel belonging to the
background. - gtThe scene without person has to be learned
before to locate people. - gtRaja et al (1998) used hue(H) and saturation(S)
of an HSI color space to decouple the influence
of illumination changes in a indoor scene, - gtA Gaussian mixture model was used to estimate
the probabilities of each pixel belonging to a
multi-colored foreground object - gtEach Gaussian models one color in the
foreground object and was learned in a training
stage.
6Motion detection in outdoor environments
illuminated by daylight
- gtA refection model influenced by ambient objects
is used. - gtlarge-scale illumination changed mainly
arises from varying cloud cover. - gtThe dominant illumination comes from either
direct sunlight or reflection from clouds - gtThe normalised rgb color space is used to
eliminate the influence of varying illumination - gtA Gaussian mixture model is used to model each
pixel of the background , provides
multi-background modelling capabilities for
complex out scenes
7Colour Fundamentals
- An image taken with a color camera is composed of
sensor responses as
-gtillumination,
-gtreflectance of an object surface
-gtCamera sensitivity
-gt wavelength,
The image intensity
gtThe appearance of objects is a result of
interaction between illumination and
reflectance. gtTo track the object surface, it is
desirable to separate the variation of the
illumination from that of the surface reflection.
8Shadow model(1)
- In an out door environment, fast illumination
changes occur at the regions where shadows emerge
or disappear. - gtlarge-scale (arising from moving cloud)
- gtsmall-scale (from objects themselves)
- Shadow model (Gershon et al 1986)
- gtThere is only one illuminant in the scene
- gtSome of the light does not reach the object
because blocking objects. - gtcreate a shadow region and a directly lit
region on the object. - gtThe shadow region is illuminated by each
reflection objects j
9Shadow model cont.
- gtThe reflected light from the object surface
- gtFor the directly lit region
gtFor the shadow region
gtAssume the chromatic average of the ambient
objects is gray i.e. it is relatively balanced in
all visible wavelengths and
Where c is independent of and may
varies over space
10Shadow model cont.
- The assumption is realistic for the fast-moving
cloud case, in which the only illuminant is the
sunlight and both the blocking and ambient
objects are gray(or white) clouds. - Under the assumption, the reflected light from
directly lit and shadow regions will stay in
proportion for a given object surface. Thus the
image intensities at all color channels being in
proportion no matter lit or shadowed - The proportionality between RGB color channels
can be represented using the normalised color
components - Where each component of will keep constant
for a given object surface under varying
illumination.
11Color-Invariant Motion Detection
- A single Gausssion is sufficient to model a pixel
value for one channel of the RGB components
resulting from a particular surface under
particular lighting and account for acquisition
noise. - A single adaptive Gaussian is sufficient to model
each RGB channel if lighting changes gradually
over time. The estimated background value is
interpolated between the previous estimation and
the new observation. It cannot follow an RGB
component under fast illumination changes. A
normalized color component (rgb) for a given
object surface tends to be constant under
lighting changes and is appropriate to model
using an adaptive Gaussion. - Multiple adaptive Gaussians (a mixture of
Gaussions) are used to model a pixel at which
multiple object surfaces may appear as the
backgrounds. E.g. swaying trees
12Color-Invariant Motion Detection cont.
- Let the pixel value at time t be
and modeled by a mixture of N
Gaussian distributions. The probability of
observing the pixel value is - Where G is the Gaussian probability density
function of the I-th background Bi, P(Xi Bi) - P(Bi) reflecting the likelihood that the
distribution accounts for the observed data. -
13Scheme
- Every new observation, Xt, is checked against the
N Gaussian distributions, A match is defined as
an observation within about 3 standard deviations
of a distribution. If none of the N distributions
match the current pixel value, the least probable
distribution is replaced by the new observation. - For the matched distribution, i, the parameters
are updated as - For the unmatched distribution
- The distribution(s) with greatest weight
is(are)considered as the background model.
14Experimental Results
- To assess the significance of the color-invariant
motion detection - gtEvaluated the model at both pixel and frame
levels using a set of image sequences. - gtThe image sequence was captured at a frame
rate of 2 hz - gtEach frame was compressed in JPEG format
- gtframe size 384x288 pixels.
- This sequence well represents the abundant
contexts of a day lit outdoor environment - gtFast illumination changes, waving trees,
shading of tree canopies, - gtHighlights of specula reflection, as well
as pedestrians.
15- the absolute (RGB) and normalised (rgb) color
components at selected pixels through time. - No foreground object is present.
- Foreground object are present.
- The absolute color components (RGB) change
greatly with the illumination. - The normalized color components (rgb) for a
background pixel have flat profiles under
illumination. - For each foreground pixel, at least one rgb
component appears as an apparent spike.
16- The parameter updating procedure of the Gaussian
background model for one color component - A lit region with foreground objects
- A shadowed region without foreground objects
- The thin lines represent and
(upper and lower) profiles, respectively.
17Comparing the RGB and rgb results under little
illumination change gtThe results are
coherent. gtBecause of the different emphasis of
image contexts, the blobs appear as
different shapes.
18- The RGB and rgb results under a major
illumination change. - A large area of the background is detected as a
huge foreground object. - (c) Ground truth targets are clearly visible
under fast illumination changes
19Discusion(1)
- gtAppropriate selection of the initial deviation
- gtAn underestimate of the initial deviation
- prohibits many ground truth
background pixels from being - adapted into background models
- gtAn overestimate of the initial deviation
- needs a longer learning period at the
start of an image sequence - gtCurrently
- it is manually selected and globally
uniform according to the - noise level in shaded regions where
the absolute noise level in - rgb components is high.
- gtIn future
- it may be automatically selected
according to the local - spatial variation in the rgb
components at the start time
20Discussion(2)
- gtThe rgbl color space
- gtcombined the intensity I, with the rgb
color space - gtis an invertible transformation from RGB
space - gtavoids the loss of the intensity
information. - gtrobustly determinates the shadowed region
- gtthe rgb components are stable
- gtthe I component is significantly
lower. - gtTwo kinds of pixels which may be excluded from
consideration - gtRGB components saturated can make the
corresponding rgb components unconstrained - gtThe rgb components in over-dark regions are
very noisy. - gtTo alleviate this problem
- gt Using cameras with auto iris control
- gt Gamma correction
21Conclutions
- gtA Gaussion mixture model based on the rgb color
space has been presented for maintaining a
background image for motion detection. - gtThe scheme is especially successful when applied
to outdoor scenes illuminated by daylight and is
robust to fast illumination changes arising from
moving cloud and self-shadows. - gtThe success results from a realistic reflection
model in which shadows are present.
22Thank you!