59 - MVA
MVA'SO
IAPR Workshop on Machine Vision Applications Nov. 28-30,1990, Tokyo
IMAGE SEQUENCE ANALYSIS OF REAL WORLD HUMAN BODY MOTION Xin ZHou,Qing LU,Zhe CU Department of Computer Science,Fudan University
220 Handan Road Shanghai,2O0433,PaR.China ABSTRACT In this paper attempts have been focused o n using window-tracking technlque to analyse human body movement.According to the behavlor of human vision systen,an Idea of rough and precise search-space is proposed in order to reduce the search-space.Some properties of windows whlch f i t tracking t h e joint m o v e m e n t s of the real world, human body are il1ustrated.Further discussion on window-tracking technique i s presented at last. INTRODUCTION To analyse human body motion ,traditional method is to indicate the l o c a t i o n s of j o i n t s in e a c h i m a g e f r a m e manually,then calculation of speeds,acceleration and trajectories 1s d o n e by c o m p u t e r . I t n e e d s g r e a t a m o u n t o f work but the result is not very precise.Another method is to put some marks on the places to be investigated,then analyse their movements and change,such a s recognition of human emotional expression done by N.Suwa and others. Besides, Kolchlro Aklta adopts a method of segment and representation under the conduction of human body model. But his model i s too simple, and the interrelationship between images in the s e q u e n c e i s not utilized, s o t h l s m e t h o d can't b e u s e d w i d e l y . In this paper we adopt a window-tracking technlque to analyse human body movement.It c a n b e used to analyse human body m o t i o n in real world.
-
shown in Fig.l,when rigid body moves from locatlon 1 to location 2,the window also t r a c k s f r o m l o c a t i o n a t o l o c a t i o n b. W e as express the technique wlth formula follow-
Fig.1 WindowFig.2 Tracking rigld body Human body model Define t h e s e c o n d f r a m e a s f(x,y), D e f i n e t h e f i r s t f r a m e a s w(x,y), t h e n t r a c k i n g e v a l u a t i o n f u n c t i o n is: O b v i o u s l y , R ( m , n ) 1 s m i n i r u m if t h e w i n d o w content of first f r a m e matches that of s e c o n d f r a m e . T h a t i s t o f i n d (mo,no)whlcn satisfied: where s is search-space. 2. H u m a n B o d y H o d e 1 : T h e h u m a n b o d y i s represented a s t h e combinatorial body of rigid bodies approximately.In this case,a joint is a unique point connecting two rigid bodies.So t h e movement of human body c a n b e d e s c r i b e d if t h e j o i n t s c a n b e tracked.Because postures of human body 1 s mainly described by 1 3 joints,and many human body movements, such a s walking and running, a r e symnetry,so we only need to track 7 j o i n t s in o u r experiment.They a r e center of head and joints o n shoulder ,elbow .wrist, hip, knee and Fig.zankle,as s h o w n i n Fig.2. 3. d c t u a l d n a l y s i s T e c h n i q u e :
UINDOU TROCKING TECHNIQUE ic
,....preproc*..,
ng ............... +
. i . + t
r a c k r n g.................. -
input
1. G e n e r a l w i n d o w - t r a c k i n g t e c h n i q u e : General wlndow-tracking technique is adopted in tracklng rigid b o d y motion. As
Fig.3 3.1 O v e r v i e w : F i g . 3 i s t h e b l o c k
.-'
output
diagram
o f t h e a n a l y s l s system.The whole p r o c e s s i s composed o f t w o s t a g e s : p r e p r o c e s s i n g and t r a c k i n g . I n f i r s t s t a g e , n o l s e i s eliminated f r o m e a c h f r a m e b y wave f i l t e r , a n d image sequence g r a y v a l u e i s u n i f l e d so t h a t e v e r y i m a g e f r a m e has t h e same a v e r a g e g r a y v a l u e . I n second stage,the window's l o c a t i o n i n p r e s e n t image f r a m e i s s o u g h t a c c o r d i n g t o t h e c o n t e n t o f t h e window i n p r e v i o u s image f r a m e , a n d t r a j e c t o r y i s shown b y tracking results. T O a n a l y s e t h e image sequenoe, first t h e j o l n t s o f t h e f i r s t lmage f r a m e a r e pointed out manually by man-machine i n t e r a c t i o n , t h e n s q u a r e windows a r e opened a t each l o c a t l o n o f j o i n t s i n f i r s t frame by computer ,so as t o t r a c k the j o i n t s i n n e x t frame .The c o n t e n t o f t h e windows changes r e l e v a n t l y d u r l n g t h e procedure. 3.2 Preprocess 1)Smoothing and Noise:We 5 Eliminate adopt fast 2 middle-value f l l t e r t o eliminate noise of inputed images.The structure of the IS shown as Fig.4 filter F1g.4 2)Gray Value Uniformatlon: As the lmages' b r i g h t n e s s o f each frame c a n ' t be absolutely the same,ltls posslble that the o b j e c t has different gray value In different f r a m e . S o i t ' s n e c c e s s a r y t o u n i f y t h e g r a y v a l u e o f e a c h image f r a m e . Suppose t h e r e a r e M f r a m e s i n a image s e q u e n c e , a n d e a c h image i s made u p o f n*m plxels(1n our experiment,n=512,m:512). fk(i,j) i s the gray value o f the k t h
technique
which
s e a r c h l o c a t i o n o f t h e window i n s u c c e e d l n g image b y m a t c h i n g t h e c o n t e n t o f t h e window i n p r e v i o u s image. That is to find c o o r d i n a t e p o i n t (mo,no) in succeedlng image frame which
satisfied
formula
(2).
The window s i z e a n d t r a c k i n g t a c t i c s s h o u l d be considered. 1 ) Window S i z e : A s t h e b a c k g r o u n d o f the image i s r a t h e r c o m p l e x and may be different in different frame,whatls more,the surface of human body is smooth,and t h e g r a y v a l u e o f t h e p i x e l s i s c l o s e t o g e t h e r , s o i f t h e window s i z e i s t o o big,and much background is enclosed ,tracking c a n ' t be p r e c l s e . 8 u t i f the window s i z e i s s o s m a l l t h a t t h e window i s i n t h e a r e a o f human b o d y , b o u n d a r i e s a n d o u t l i n e s which cause b l g d i f f e r e n c e o f g r a y v a l u e w i l l 1ose.We g e t t w o rules by e n p e r l m e n t . T h e wlndow s h o u l d i n c l u d e more o u t l i n e s and b o u n d a r i e s s o t h a t t h e g r a y has much h i s t o g r a m o f wlndow c o n t e n t d i f f e r e n c e . A t t h e mean t i m e , t h e window should enclosed l e s s background.In t a c t the to r u l e s a r e c o n t r a d l c t o r y , ~ ~we h a v e conslder trade-off when we c h o o s e the window s i z e . We s e t d i f f e r e n t window o n d i f f e r e n t j o i n t considering i t s d l s t i n c t f e a t u r e s . I n c h a r t . l , L x , L y e a c h s t a n d s f o r t h e window s l z e on x - a x i s and y - a x i s . C h a r t . 1 The Window S i z e f o r D e t t e r r n t J o i n t
...m;
image,where k = O , l , . . . M-1; 1=1,2 ~ = 1 , 2 n. Then t h e a v e r a g e g r a y v a l u e each frame i s :
...
of
-
l m n Z , Z f (i,j) a v g k - mn i = i J = ik The s e c o n d a n d t h e next Images are processed as f o l l o w : s t e p 1 : f ( i , j ) = f ( i , j ) + avg -avg k k 0 k where x i s t h e t a k e - d o w n I n t e g e r s t e p 2: 1 . Generating random number o I n (0,l)
-
2.
3.3 T r a c k i n g Window-tracking i s the
ft(l.~)+l
if
05 [ a v g - a v g ] 0
where [ x ] = x - x
;k=1,2
. . .M-1.
k
2 ) T r a c k i n g T a c t l c s : 0s t h e background o f human body m o t l o n images i s r e a l w o r l d , p u r e t r a c k l n g m e t h o d s s u c h a s g r a d i e n t methods c a n ' t be u s e d i n t h i s case.And because complex background may cause several minlmum v a l u e s i n search space,global search 1s neccessary.Based on i t , t h e l e a s t i s found as the r e s u l t o f tracking. There a r e two c o n s t r a i n t s t o reduce t h e s e a r c h space. D i s t a n c e Constra1nt:The displacement of a between p o i n t on the motive 4 o b j e c t
succeeding images 1s d e f i n l t e , b e c a u s e the i n t e r v a l 7 between two frame i s v e r y s h o r t . Constraint:flov~ng object's Velocity v e l o c i t y w o n ' t change much l n 7 a s a r e s u l t interla.So the range o f a polnt's of r e l e v a n t l o c a t l o n i n s u c c e e d i n g image c a n be e s t i m a t e d a c c o r d i n g t o the p r e v l o u s v e l o c l t y o f the o b j e c t . Suppose f i n i s h e d s e q u e n t f r a m e s are f , p r e s e n t f r a m e i s fk+l; k-1' k moving object has
A point
f
( X - ~ , Vl ) , ( ~ o , ~ o )
and ( x , v )
f
separate1y;Dlstance k+l ,velocity c o n s t r a i n t 1 s
space s
h
S
s h 2= ( ( x , v )
and is
t
d search
.Then
I
1
2
2
2
(x-xO) t ( Y - Y ) 5 T ) 2 O 2 d 2 ( X - ~ X ~ ~ +X( -v ~- ~) v ~ + v -5~ 1 )v
When people want to find an object,usually they search i t s l o c a t i o n roughly i n a b i g space,then take a f u r t h e r s t e p t o decide t h e p r e c i s e 1ocation.We accept thls idea in searching t e c h n i q u e , a d o p t a two-step method t o reduce t h e s e a r c h space. S t e p 1:Rough S e a r c h a n d M a t c h Rough l o c a t i o n c a n b e obtained by s c a n n i n g t h e image o n i n t e r l a c i n g c o l u m n a n d row.So rough s e a r c h space can be d e f i n e d as f o l l o w :
So r o u g h m a t c h i n g 1 s : min R(m,n)= R(k,l)
tk , l ) € S
1
i s r a t h e r s m a l 1 , t h e n urn- , t h a t h 2 i s , search time can be reduced t o n e a r l y h a l f o f g l o b a l search time.
gene rally,^
CONCLUSION
: ((x,v)
hi
1t8
a
on
i s
S = S ns h hl h2
where
v
t x , y)Es
coordinate
fk-l,fk
on
constraint
t
c o s t b y g l o b a l s e a r c h and m a t c h mathod. s u p p o s e one o p e r a t i o n o f s e a r c h a n d two match spends t i m e t , t h e n t h e r a t l o o f methods' t i m e consumtion i s :
a
Step 2:Precise Search and Hatch: Precise search i s c o r r e c t i n g the o f rough search.Suppose location given by rough ( m , n ) , t h e n p r e c i s e s e a r c h space
the r e s u l t approximate saerch is S i s made
P
u p o f (m,n) a n d i t s n e a r e s t p i x e l s . s = { ( x , v ) I ( (x-m J I l ) A ( I Y - n 1 5 1 ) )
P
S o , p r e c i s e s e a r c h and match i s
I n t h e s e c t ~ o nf o l l o w s , w e compare t h e t i m e c o s t b y t w o - s t e p s e a r c h ardd m a t c h w i t h t h a t
H e r e we r e v i e w some p r o b l e m s t h a t require futher discussion. The t e c h n i q u e c a n b e i m p r o v e d i n t w o respects.First,the major content o f the window i s a p a r t o f the object being tracked,and the s u r f a c e o f the object(human b o d y p a r t s ) 1 s r a t h e r smooth.So generally the gray histogram w i l l appear as Fig.5,where t h e r e a r e I - - n frequency peaks w h i c h a r e o b v i o u s l y different f r o m t h e others. Their correspondent gray value s e c t i o n i s t h e range o f t h e gray v a l u e o f tracked.So,we can the object being e l i m i n a t e b a c k g r o u n d i n t h e window by mapping o t h e r g r a y v a l u e t o a s i n g l e g r a y v a l u e . S e c o n d , t h e window t r a c k i n g t e c h n i q u e w i l l f i t f o r t r a c k i n g t h e movement w h i c h changes i n t h e d i r e c t i o n o f d e p t h by a d d i n g scale factor. B u t s t i l l t h e r e a r e some p r o b l e m s . T h e f i r s t one i s t h a t i m a g e s s h o u l d n o t have much n o i s e b e c a u s e t r a c k i n g a n d m a t c h i n g i s done i n t h e l i g h t o f g r a y v a l u e , o t h e r w i s e t h e r e s u l t s won't be p r e c i s e . Another track p r o b l e m i s t h a t t h i s technique c a n ' t t h e o b j e c t i n v i s i b l e i n some f r a m e s o f the sequence.
ACKNOWLEDGHENT
T h e s u p p o r t of t h e N a t i o n a l for Natural Sience is acknowledged.
Foundation gratefully
REFERENCES [l]
N.Suna,N.Sugie and K.FuJlmurr,"A P r e l l m l n a r y N a t e o n P a t t e r n Recognition of H u m a n Emotional E x p r e s s i o n " , P r o c . o f t h e 4 t h I J C P R , 1 9 7 8 , pp408-410. [2] K.Aklta,"Image S e q u e n c e A n a l y s i s of R e a l W o r l d W o m a n M o t l o n " , P.R, Vo1.17, No.1, 1 9 8 4 , pp73-83. [3] T.S.Huang a n d R.Y.Tsal,"Image Sequence Motlon Estlmatlon",Image Analysis: Sequence Analysis (ed. by T.S.Huang),Sprlng-Verlag Berlin, 1981, ppl-36. [4] S . U l l m a n , " R e c e n t C o m p u t a t i o n a l S t u d l e s in t h e Interpretation of S t r u c t u r e f r o m Motlon",Human and Machlne Vlsion,tlIT Press, Cambridge, Masschusetts, 1981, pp459-479. [5]M.K.Leung a n d Y.H.Yang "Human Body M o t l o n Segmentation In a C o m p l e x S c e n e , " P.R., Vol.20, N o . 1 , 1 9 8 7 , p p 5 5 - 6 4 . [6]M.K.Leung a n d Y.H.Yang,"A Reglon Based Approach for Human Body flotlon Analysis," P.R., Vo1.20, N0.3, 1987,pp321-339. 7
IAPR Workshop on Machine Vision Applications Nov. 28-30,1990, Tokyo
IMAGE SEQUENCE ANALYSIS OF REAL WORLD HUMAN BODY MOTION Xin ZHou,Qing LU,Zhe CU Department of Computer Science,Fudan University
220 Handan Road Shanghai,2O0433,PaR.China ABSTRACT In this paper attempts have been focused o n using window-tracking technlque to analyse human body movement.According to the behavlor of human vision systen,an Idea of rough and precise search-space is proposed in order to reduce the search-space.Some properties of windows whlch f i t tracking t h e joint m o v e m e n t s of the real world, human body are il1ustrated.Further discussion on window-tracking technique i s presented at last. INTRODUCTION To analyse human body motion ,traditional method is to indicate the l o c a t i o n s of j o i n t s in e a c h i m a g e f r a m e manually,then calculation of speeds,acceleration and trajectories 1s d o n e by c o m p u t e r . I t n e e d s g r e a t a m o u n t o f work but the result is not very precise.Another method is to put some marks on the places to be investigated,then analyse their movements and change,such a s recognition of human emotional expression done by N.Suwa and others. Besides, Kolchlro Aklta adopts a method of segment and representation under the conduction of human body model. But his model i s too simple, and the interrelationship between images in the s e q u e n c e i s not utilized, s o t h l s m e t h o d can't b e u s e d w i d e l y . In this paper we adopt a window-tracking technlque to analyse human body movement.It c a n b e used to analyse human body m o t i o n in real world.
-
shown in Fig.l,when rigid body moves from locatlon 1 to location 2,the window also t r a c k s f r o m l o c a t i o n a t o l o c a t i o n b. W e as express the technique wlth formula follow-
Fig.1 WindowFig.2 Tracking rigld body Human body model Define t h e s e c o n d f r a m e a s f(x,y), D e f i n e t h e f i r s t f r a m e a s w(x,y), t h e n t r a c k i n g e v a l u a t i o n f u n c t i o n is: O b v i o u s l y , R ( m , n ) 1 s m i n i r u m if t h e w i n d o w content of first f r a m e matches that of s e c o n d f r a m e . T h a t i s t o f i n d (mo,no)whlcn satisfied: where s is search-space. 2. H u m a n B o d y H o d e 1 : T h e h u m a n b o d y i s represented a s t h e combinatorial body of rigid bodies approximately.In this case,a joint is a unique point connecting two rigid bodies.So t h e movement of human body c a n b e d e s c r i b e d if t h e j o i n t s c a n b e tracked.Because postures of human body 1 s mainly described by 1 3 joints,and many human body movements, such a s walking and running, a r e symnetry,so we only need to track 7 j o i n t s in o u r experiment.They a r e center of head and joints o n shoulder ,elbow .wrist, hip, knee and Fig.zankle,as s h o w n i n Fig.2. 3. d c t u a l d n a l y s i s T e c h n i q u e :
UINDOU TROCKING TECHNIQUE ic
,....preproc*..,
ng ............... +
. i . + t
r a c k r n g.................. -
input
1. G e n e r a l w i n d o w - t r a c k i n g t e c h n i q u e : General wlndow-tracking technique is adopted in tracklng rigid b o d y motion. As
Fig.3 3.1 O v e r v i e w : F i g . 3 i s t h e b l o c k
.-'
output
diagram
o f t h e a n a l y s l s system.The whole p r o c e s s i s composed o f t w o s t a g e s : p r e p r o c e s s i n g and t r a c k i n g . I n f i r s t s t a g e , n o l s e i s eliminated f r o m e a c h f r a m e b y wave f i l t e r , a n d image sequence g r a y v a l u e i s u n i f l e d so t h a t e v e r y i m a g e f r a m e has t h e same a v e r a g e g r a y v a l u e . I n second stage,the window's l o c a t i o n i n p r e s e n t image f r a m e i s s o u g h t a c c o r d i n g t o t h e c o n t e n t o f t h e window i n p r e v i o u s image f r a m e , a n d t r a j e c t o r y i s shown b y tracking results. T O a n a l y s e t h e image sequenoe, first t h e j o l n t s o f t h e f i r s t lmage f r a m e a r e pointed out manually by man-machine i n t e r a c t i o n , t h e n s q u a r e windows a r e opened a t each l o c a t l o n o f j o i n t s i n f i r s t frame by computer ,so as t o t r a c k the j o i n t s i n n e x t frame .The c o n t e n t o f t h e windows changes r e l e v a n t l y d u r l n g t h e procedure. 3.2 Preprocess 1)Smoothing and Noise:We 5 Eliminate adopt fast 2 middle-value f l l t e r t o eliminate noise of inputed images.The structure of the IS shown as Fig.4 filter F1g.4 2)Gray Value Uniformatlon: As the lmages' b r i g h t n e s s o f each frame c a n ' t be absolutely the same,ltls posslble that the o b j e c t has different gray value In different f r a m e . S o i t ' s n e c c e s s a r y t o u n i f y t h e g r a y v a l u e o f e a c h image f r a m e . Suppose t h e r e a r e M f r a m e s i n a image s e q u e n c e , a n d e a c h image i s made u p o f n*m plxels(1n our experiment,n=512,m:512). fk(i,j) i s the gray value o f the k t h
technique
which
s e a r c h l o c a t i o n o f t h e window i n s u c c e e d l n g image b y m a t c h i n g t h e c o n t e n t o f t h e window i n p r e v i o u s image. That is to find c o o r d i n a t e p o i n t (mo,no) in succeedlng image frame which
satisfied
formula
(2).
The window s i z e a n d t r a c k i n g t a c t i c s s h o u l d be considered. 1 ) Window S i z e : A s t h e b a c k g r o u n d o f the image i s r a t h e r c o m p l e x and may be different in different frame,whatls more,the surface of human body is smooth,and t h e g r a y v a l u e o f t h e p i x e l s i s c l o s e t o g e t h e r , s o i f t h e window s i z e i s t o o big,and much background is enclosed ,tracking c a n ' t be p r e c l s e . 8 u t i f the window s i z e i s s o s m a l l t h a t t h e window i s i n t h e a r e a o f human b o d y , b o u n d a r i e s a n d o u t l i n e s which cause b l g d i f f e r e n c e o f g r a y v a l u e w i l l 1ose.We g e t t w o rules by e n p e r l m e n t . T h e wlndow s h o u l d i n c l u d e more o u t l i n e s and b o u n d a r i e s s o t h a t t h e g r a y has much h i s t o g r a m o f wlndow c o n t e n t d i f f e r e n c e . A t t h e mean t i m e , t h e window should enclosed l e s s background.In t a c t the to r u l e s a r e c o n t r a d l c t o r y , ~ ~we h a v e conslder trade-off when we c h o o s e the window s i z e . We s e t d i f f e r e n t window o n d i f f e r e n t j o i n t considering i t s d l s t i n c t f e a t u r e s . I n c h a r t . l , L x , L y e a c h s t a n d s f o r t h e window s l z e on x - a x i s and y - a x i s . C h a r t . 1 The Window S i z e f o r D e t t e r r n t J o i n t
...m;
image,where k = O , l , . . . M-1; 1=1,2 ~ = 1 , 2 n. Then t h e a v e r a g e g r a y v a l u e each frame i s :
...
of
-
l m n Z , Z f (i,j) a v g k - mn i = i J = ik The s e c o n d a n d t h e next Images are processed as f o l l o w : s t e p 1 : f ( i , j ) = f ( i , j ) + avg -avg k k 0 k where x i s t h e t a k e - d o w n I n t e g e r s t e p 2: 1 . Generating random number o I n (0,l)
-
2.
3.3 T r a c k i n g Window-tracking i s the
ft(l.~)+l
if
05 [ a v g - a v g ] 0
where [ x ] = x - x
;k=1,2
. . .M-1.
k
2 ) T r a c k i n g T a c t l c s : 0s t h e background o f human body m o t l o n images i s r e a l w o r l d , p u r e t r a c k l n g m e t h o d s s u c h a s g r a d i e n t methods c a n ' t be u s e d i n t h i s case.And because complex background may cause several minlmum v a l u e s i n search space,global search 1s neccessary.Based on i t , t h e l e a s t i s found as the r e s u l t o f tracking. There a r e two c o n s t r a i n t s t o reduce t h e s e a r c h space. D i s t a n c e Constra1nt:The displacement of a between p o i n t on the motive 4 o b j e c t
succeeding images 1s d e f i n l t e , b e c a u s e the i n t e r v a l 7 between two frame i s v e r y s h o r t . Constraint:flov~ng object's Velocity v e l o c i t y w o n ' t change much l n 7 a s a r e s u l t interla.So the range o f a polnt's of r e l e v a n t l o c a t l o n i n s u c c e e d i n g image c a n be e s t i m a t e d a c c o r d i n g t o the p r e v l o u s v e l o c l t y o f the o b j e c t . Suppose f i n i s h e d s e q u e n t f r a m e s are f , p r e s e n t f r a m e i s fk+l; k-1' k moving object has
A point
f
( X - ~ , Vl ) , ( ~ o , ~ o )
and ( x , v )
f
separate1y;Dlstance k+l ,velocity c o n s t r a i n t 1 s
space s
h
S
s h 2= ( ( x , v )
and is
t
d search
.Then
I
1
2
2
2
(x-xO) t ( Y - Y ) 5 T ) 2 O 2 d 2 ( X - ~ X ~ ~ +X( -v ~- ~) v ~ + v -5~ 1 )v
When people want to find an object,usually they search i t s l o c a t i o n roughly i n a b i g space,then take a f u r t h e r s t e p t o decide t h e p r e c i s e 1ocation.We accept thls idea in searching t e c h n i q u e , a d o p t a two-step method t o reduce t h e s e a r c h space. S t e p 1:Rough S e a r c h a n d M a t c h Rough l o c a t i o n c a n b e obtained by s c a n n i n g t h e image o n i n t e r l a c i n g c o l u m n a n d row.So rough s e a r c h space can be d e f i n e d as f o l l o w :
So r o u g h m a t c h i n g 1 s : min R(m,n)= R(k,l)
tk , l ) € S
1
i s r a t h e r s m a l 1 , t h e n urn- , t h a t h 2 i s , search time can be reduced t o n e a r l y h a l f o f g l o b a l search time.
gene rally,^
CONCLUSION
: ((x,v)
hi
1t8
a
on
i s
S = S ns h hl h2
where
v
t x , y)Es
coordinate
fk-l,fk
on
constraint
t
c o s t b y g l o b a l s e a r c h and m a t c h mathod. s u p p o s e one o p e r a t i o n o f s e a r c h a n d two match spends t i m e t , t h e n t h e r a t l o o f methods' t i m e consumtion i s :
a
Step 2:Precise Search and Hatch: Precise search i s c o r r e c t i n g the o f rough search.Suppose location given by rough ( m , n ) , t h e n p r e c i s e s e a r c h space
the r e s u l t approximate saerch is S i s made
P
u p o f (m,n) a n d i t s n e a r e s t p i x e l s . s = { ( x , v ) I ( (x-m J I l ) A ( I Y - n 1 5 1 ) )
P
S o , p r e c i s e s e a r c h and match i s
I n t h e s e c t ~ o nf o l l o w s , w e compare t h e t i m e c o s t b y t w o - s t e p s e a r c h ardd m a t c h w i t h t h a t
H e r e we r e v i e w some p r o b l e m s t h a t require futher discussion. The t e c h n i q u e c a n b e i m p r o v e d i n t w o respects.First,the major content o f the window i s a p a r t o f the object being tracked,and the s u r f a c e o f the object(human b o d y p a r t s ) 1 s r a t h e r smooth.So generally the gray histogram w i l l appear as Fig.5,where t h e r e a r e I - - n frequency peaks w h i c h a r e o b v i o u s l y different f r o m t h e others. Their correspondent gray value s e c t i o n i s t h e range o f t h e gray v a l u e o f tracked.So,we can the object being e l i m i n a t e b a c k g r o u n d i n t h e window by mapping o t h e r g r a y v a l u e t o a s i n g l e g r a y v a l u e . S e c o n d , t h e window t r a c k i n g t e c h n i q u e w i l l f i t f o r t r a c k i n g t h e movement w h i c h changes i n t h e d i r e c t i o n o f d e p t h by a d d i n g scale factor. B u t s t i l l t h e r e a r e some p r o b l e m s . T h e f i r s t one i s t h a t i m a g e s s h o u l d n o t have much n o i s e b e c a u s e t r a c k i n g a n d m a t c h i n g i s done i n t h e l i g h t o f g r a y v a l u e , o t h e r w i s e t h e r e s u l t s won't be p r e c i s e . Another track p r o b l e m i s t h a t t h i s technique c a n ' t t h e o b j e c t i n v i s i b l e i n some f r a m e s o f the sequence.
ACKNOWLEDGHENT
T h e s u p p o r t of t h e N a t i o n a l for Natural Sience is acknowledged.
Foundation gratefully
REFERENCES [l]
N.Suna,N.Sugie and K.FuJlmurr,"A P r e l l m l n a r y N a t e o n P a t t e r n Recognition of H u m a n Emotional E x p r e s s i o n " , P r o c . o f t h e 4 t h I J C P R , 1 9 7 8 , pp408-410. [2] K.Aklta,"Image S e q u e n c e A n a l y s i s of R e a l W o r l d W o m a n M o t l o n " , P.R, Vo1.17, No.1, 1 9 8 4 , pp73-83. [3] T.S.Huang a n d R.Y.Tsal,"Image Sequence Motlon Estlmatlon",Image Analysis: Sequence Analysis (ed. by T.S.Huang),Sprlng-Verlag Berlin, 1981, ppl-36. [4] S . U l l m a n , " R e c e n t C o m p u t a t i o n a l S t u d l e s in t h e Interpretation of S t r u c t u r e f r o m Motlon",Human and Machlne Vlsion,tlIT Press, Cambridge, Masschusetts, 1981, pp459-479. [5]M.K.Leung a n d Y.H.Yang "Human Body M o t l o n Segmentation In a C o m p l e x S c e n e , " P.R., Vol.20, N o . 1 , 1 9 8 7 , p p 5 5 - 6 4 . [6]M.K.Leung a n d Y.H.Yang,"A Reglon Based Approach for Human Body flotlon Analysis," P.R., Vo1.20, N0.3, 1987,pp321-339. 7
