DETECTION OF A MOVING OBJECT IN THE IMAGE463
velocity of the object based on the observed realizationand to detect using the maximum-likelihood method[12, 13].
4. THE CHARACTERISTICS OF THE MAXIMUM-LIKELIHOOD DETECTORWhen detecting a uniformly colored object (s(r) =s0) against a homogeneous background (ν(r) = ν0), the
(V0,maximum value of the signal function is z2 = S
V0) = 2(s0 – ν0)2TGs/N0, where Gs is the area of theobject image. Then, the normalized signal function canbe written in the form
1
S(V,V0)=---S(V,V0)2z
1= ---------GsT
T
piecewise boundary such that every straight line inter-sects it at no more than two points.
Let the segment [ymin, ymax] of the axis Y be the pro-jection of the domain s (together with the boundary Γ)onto this axis (Fig. 2). Let us draw through the pointsymin and ymax straight lines parallel to the axis X as isshown in Fig. 2. Each of these straight lines intersectsthe boundary Γ at only one point. These two intersec-tion points of the speci ed straight lines with theboundary Γ separate it into two curves Γ– and Γ+, whichcan uniquely be projected on the axis Y.
Let us denote by lV = ymax – ymin the length of the pro-jection of the domain s on the axis Y, i.e., on the direc-tion perpendicular to the direction of motion. Let usdivide the segment [ymin, ymax] into n equal parts. Forthis purpose, we draw n – 1 straight lines parallel to theaxis X so that the distances between the adjacentstraight lines are the same and are h = lV/n. The seg-ments of the adjacent straight lines that lie in thedomain D, i.e., con ned between the left boundaries ofthe domains given by the indicator functions Is(x, y) andIs(x – | V|t, y), have the same length | V|t. When n∞ and | V|, the curvilinear trapezoids generatedby these segments and the boundaries of the domainscon ned by these segments can be regarded as parallel-ograms with bases | V|t and altitudes h. Taking intoaccount that the area of each parallelogram is S1 =| V|t h = | V|tlV/n and that we have n parallelograms,we nd that J(t) ≈ n S1 = | V|tlV. Here, we ignore theareas of the curvilinear triangles con ned by the bound-aries of the objects and straight lines y = ymax and y =ymin since these areas have the infinitesimal order V2.Thus, when | V|, the signal function can be rep-y(4.1)
s
s
∫∫∫I(x–Vt,y)I(x–Vt,y)dxdydt.
0
Introducing the notation V = V – V0, we have
S(V,V0)=S(V–V0)=S( V)1
= ---------GsT
T
∫∫∫I(x,y)I(x– Vt,y)dxdydt.
s
s
0
2
Taking into account that Is(x, y) = Is(x, y), we rewritethe signal function as
S(V–V0)=S( V)=1
1–---------GsT
T
∫∫∫I(x,y)[I(x,y)–I(x– Vt,y)]dxdydt.
s
s
s
0
Let us investigate the behavior of the signal functionwhen | V|0. The quantityJ(t)=
∫∫
Is(x,y)[Is(x,y)–Is(x– Vt,y)]dxdy
=
∫∫[I(x,

y)–I(x– Vt,y)]dxdy
s
s
s
at the moment t is the area of the gure D (in general,multiply-connected) that completely belongs to s,none of whose point belongs to the domain where Is(x –| V|t, y) = 1.
In what follows, we assume that s is a simply con-nected nite domain with a piecewise smooth boundaryΓ and that any straight line parallel to the motion direc-tion intersects the boundary Γ at no more than twopoints. We call the domains of this type the domain oftype KV. Note that D is also a domain of type KV. Theclass KV extends the class of domains of type K [16],which contains simply connected nite domains with a
yFig. 2.
JOURNAL OF COMPUTER AND SYSTEMS SCIENCES INTERNATIONAL Vol. 45 No. 3 2006
NorthChina Univ.ofTechnology,Beljing 100041China) Abstract:Thefordetection....beusedto打ackmoving object, fordigitalimageprocessing discussed. character,...
Detection of a moving ob... 10页 5财富值 A Moving Object Detectio......In the image model Eq. 1, radial photometric distortions of perspective ...
Moving Object Detection and Tracking in Video Based on Higher Order ...to the wavelet transform coecients of the motion compensated image frames....
1, NO. 1, MARCH 1999 65 Detection of Moving Cast Shadows for Object ...In the image model (1), radial photometric distortions of perspective ...
motion and geometry based obstacle detection in image sequences of traffic ...Up to now the classi cation whether a (relatively) moving object is a ...
REAL-TIME MOVING OBJECT DETECTION IN VIDEO SEQUENCES USING SPATIO-TEMPORAL ...foreground objects from the background a pure background image has to be ...
of a Moving Object from Range Video Farzad ...in the range image, and a rigid transform is ...A blob detection algorithm based on the target ...
some extensions to the method for moving object detection presented in [4]...the imaging surface) from spatio-temporal patterns of image intensity [1]....
Detection of a moving ob... 10页 5财富值 Object detection and tra......Their effectiveness in image processing still remains to confirm. The discrete...
be detected as foreground object and plauge the moving objects segmentation. ...shadow detection algorithm can greatly benefit the practical image analysis ...
热门文档