Edge Finding, Segmentation Of Edges And Recognition ... - CiteSeerX

EDGE F I N D I N G , SEGMENTATION OF EDGES AND RECOGNITION OF COMPLEX OBJECTS Yoshiaki S h i r a i E l e c t r o t c c h n i c a l Laboratory Tokyo, Japan ABSTRACT T h i s paper d e s c r i b e s a n approach t o t h e r e c o g n i t i o n of r e a l - w o r l d o b j e c t s such as books or a t e l e p h o n e o n a d e s k . The s y s t e m c o n s i s t s o f ( l ) edge f i n d i n g p r o c e s s w h i c h e x t r a c t s edges o f curved o b j e c t s from l i g h t i n t e n s i t y data, (2) s e g m e n t a t i o n o f t h e edges i n t o s t r a i g h t l i n e s o r e l l i p t i c curves, (3) r e c o g n i t i o n of objects by m a t c h i n g t h e l i n e r , t o t h e m o d e l s , and ( 4 ) s i m p l e supervisor. The module ( 1 ) , ( 2 ) and ( 3 ) i n t e r a c t w i t h each o t h e r t h r o u g h a s i m p l e s u p e r v i s o r s o t h a t t h e s y s t e m can l o c a t e g i v e n o b j e c t s q u i c k l y F i r s t , most r e l i a b l e edges a r e f o u n d and s e g m e n t e d , and t h e n r e c o g n i t i o n i r. a t t e m p t e d u s i n g segmented lines. I f r e c o g n i t i o n i s not s u c c e s s f u l , less r e l i a b l e edges a r e s e a r c h e d f o r , arid r e c o g n i t i o n i s r e t r i e d . A n e x a m p l e o f l o c a t i n g a l a m p , a book s t a n d and a t e l e p h o n e i s shown. 1.

INTRODUCTION

T h i s paper d e s c r i b e s a n approach t o t h e r e c o g n i t i o n o f r e a l - w o r l d o b j e c t s s u c h a s books o r a t e l e p h o n e o n a d e s k . Scene a n a l y s i s c o u l d b e d i v i d e d i n t o two modules: ( 1 ) e x t r a c t i o n o f f e a t u r e s and ( ? ) r e c o g n i t i o n o f o b j e c t s based o n the f e a t u r e s . I t i s o f t e n p o i n t e d o u t t h a t t h e two modules s h o u l d n o t b e s i m p l y c o n n e c t e d s e r i a l l y , b u t s h o u l d i n t e r a c t w i t h each o t h e r t o a c h i e v e reliable recognition [1]. The a u t h o r made a n a t t e m p t t o r e c o g n i s e p o l y h e d r a b y t h o s e two modules i . e . a l i n e f i n d e r a n d a l i n e p r o p o s e r [ ? ] . T h e r e h a v e b e e n many works o n m o d e l - d r i v e n r e c o g n i t i o n o f p o l y h e d r a [ 3 , 4 . 5 ] . However, i t i s n o t e a s y t o make t h e s e t w o modules c a p a b l e o f r e c o g n i z i n g c u r v e d o b j e c t s . The d i f f i c u l t p r o b l e m i s t o d e t e r m i n e w h a t f e a t u r e s a r e u s e f u l f o r r e c o g n i t i o n and how t o I n t e r p r e t ? the e x t r a c t e d f e a t u r e s . T h e r e have b e e n some w o r k s o n d e s c r i p t i o n o f curved o b j e c t s u s i n g range data [ 6 , 7 , 8 ] Garvey a n d Tenenbaum [ 9 ] made an e x p e r i m e n t f o r l o c a t i n g o b j e c t s i n o f f i c e scenes u s i n g r a n g e and c o l o r data. The w o r k p r e s e n t e d h e r e u s e s o n l y l i g h t i n t e n s i t y d a t a , because i t can e a s i l y b e o b t a i n e d and s t o r e d , w h i c h makes i t p r a c t i c a l t o a p p r o x i m a t e unknown s c e n e s a t f i r s t . Features of o b j e c t s to be e x t r a c t e d f o r recogn i t i o n d e p e n d on t h e o b j e c t s . F o r o u t d o o r scenes, r e g i o n s o f u n i f o r m c o l o r m i g h t b e good clues f o r r e c o g n i t i o n o f the sky, f i e l d s , lakes and s o o n [ 1 0 , 1 1 ] . For e x a m p l e , a l a r g e b l u e r e g i o n a t a n u p p e r p a r t o f a scene m i g h t b e t h e sky. For o f f i c e s c e n e s composed o f known o b j e c t s [ 9 ] , c o l o r and r a n g e o f t h e e a c h p o i n t i n t h e scene m i g h t b e enough t o l o c a t e t h e t a b l e t o p o r t h e d o o r candidates. O n t h e o t h e r h a n d , r e c o g n i t i o n o f e v e r y day o b j e c t s o n t h e desk u s i n g o n l y l i g h t i n t e n s i t y d a t a r e q u i r e s more c o m p l e x f e a t u r e s . T h e r e i s l e s s constraint on the c o n f i g u r a t i o n of objects on the desk t h a n t h o s e i n t h e o u t d o o r s c e n e s . IAght i n t e n s i t y o f a p o i n t i n t h e unknown d e s k s c e n e a l o n e can n o t s u g g e s t any m e a n i n g f u l f a c t s . It m i g h t b e n e c e s s a r y t o l o o k a t t h e change o f l i g h t i n t e n s i t y and t h e r e l a t i o n w i t h o t h e r p a r t s o f t h e

674

scene. T h e r e h a v e b e e n many w o r k s o n f e a t u r e e x t r a c t i o n o f complex scenes [ 1 2 , 1 3 ] , b u t few works have r e c o g n i z e d t h e s c e n e s , u s i n g t h e s e f e a t u r e s . W h i l e , many p a p e r s a s s u m i n g p i c t u r e p r i m i t i v e s , have d e a l t w i t h t h e s y m b o l i c p r i m i t i v e s t o r e c o g n i z e c o m p l e x o b j e c t s . T h e r e h a v e b e e n few w o r k s o n r e c o g n i t i o n o f complex o b j e c t s u s i n g raw l i g h t i n t e n s i t y data. The a u t h o r has p r o p o s e d l o c a l f e a t u r e s and g l o b a l features o f curved o b j e c t s f o r a n a l y s i s o f r e a l - w o r l d scenes [ l ^ ] . The l o c a l f e a t u r e s r e p r e s e n t t h e t y p e s o f 1 i g h t i n t e n s i t y change and t h e g l o b a l ones a r e t h e c o n n e c t e d l o c a l f e a t u r e s o f t h e same t y p e . I f t h e c u r v e d o b j e c t has u n i f o r m surfaces, the g l o b a l features correspond to the edges o f t h e p l a n e o r c u r v e d s u r f a c e s . T s u j i [ 1 5 ] made a l i n g u i s t i c a p p r o a c h t o s e g m e n t a t i o n o f c u r v e d o b j e c t scenes i n t o r e g i o n s using p i c t u r e p r i m i t i v e s s i m i l a r t o the l o c a l f e a t u r e s d e s c r i b e d above. It consists of hiera r c h i c a l modules, each o f w h i c h compresses t h e i n p u t d a t a t o f i n a l l y o b t a i n t h e p l a n e and c u r v e d regions. The p r o b l e m w i t h t h e method i s t h a t i t must c o m p l e t e l y a n a l y z e t h e w h o l e scene b e f o r e i t extract?', some u s e f u l f a c t s a b o u t a s c e n e . T h i s p a p e r i s a f u r t h e r e x t e n s i o n o f [ 1 4 ] and d e s c r i b e s a t o t a l system. The s y s t e m c o n s i s t s o f ( 1 ) edge f i n d i n g , ( 2 ) s e g m e n t a t i o n o f t h e edges i n t o s t r a i g h t l i n e s or e l l i p t i c curves, (3) reco g n i t i o n of objects by matching the l i n e s to the models , or by u s i r i g t h e raw d a t a i f n e c e s s a r y , and (4) top l e v e l s u p e r v i s i n g . The module ( 1 ) , ( 2 ) and < 3) a r e c o n t r o l l e d by a s i m p l e s u p e r v i s o r to i n t e r a c t w i t h each o t h e r s o t h a t d e p e n d i n g o n t h e s i t u a t i o n , t h e s y s t e m can l o c a t e g i v e n o b l e r t s q u i c k l y o r a n a l y z e t h e w h o l e s c e n e . The m o d u l e ( ] ) and ( ? ) a r e e a s i l y c o n t r o l l e d b y a s e t o f g l o b a l p a r a m e t e r s s u c h a s r e l i a b i l i t y l e v e l , minimum curvature , s p a t i a l r e s o l u t i o n , etc . They a r e a p p l i c a b l e to various objects. The module ( 3 ) w h i c h has a t p r e s e n t l e s s g e n e r a l i t y i s d e s c r i b e d b r i e f l y i n t h i s paper. P. ?.l.

EDGE FINDING

Edge p o i n t d e t e c t i o n

A n edge o f a n o b j e c t s u r f a c e i s d e f i n e d h e r e a s a s e r i e s o f s m o o t h l y c o n n e c t e d edge p o i n t s . Many o p e r a t o r s f o r edge p o i n t e x t r a c t i o n have been p r o p o s e d [ 1 2 , 1 3 , 1 6 ] . The t w o o p e r a t o r s ( t w o d i m e n s i o n a l and o n e - d i m e n s i o n a l ) d e s c r i b e d h e r e a r e d e s i g n e d t o d e t e c t edge p o i n t s o f p l a n e s and smoothly curved surfaces. The o p e r a t o r s e x t r a c t d i s c o n t i n u i t y o f b o t h t h e l i g h t i n t e n s i t y and i t s g r a d i e n t . I f t h e d i r e c t i o n o f t h e edge i s unknown, t w o - d i m e n s i o n a l o p e r a t o r i s u s e d t o compute t h e d i r e c t i o n o f t h e g r a d i e n t , o t h e r w i s e a oned i m e n s i o n a l o p e r a t o r d e t e c t s t h e edge p o i n t qui c k l y . The f o r m e r o p e r a t o r makes use o f t h e g r a d i e n t computed by a c o n v e n t i o n a l method [14 One o f t h r e e r e g i o n s shown i n F i g . 1 i n used t o g e t t h e g r a d i e n t at the c e n t e r element. The g r a d i e n t v a l u e D a n d t h e d i r e c t i o n a f o r mode 1 a r e d e f i n e d b y l i g h t i n t e n s i t y o f the neighboring points I ( i ) a s

(C1 ) meanr, that the light intensity at an edge must change. (C2) means that the neighbor of an edge i n direction a must have a different l i g h t intensity change from that at the edge. Usually if (C?) holds for both i 1 >0 and ip^O, then the edge corresponds to the boundary between two planes. If only one side of the edge (i0) satisfies (C?), it corresponds to an edge of a curved surface. Fdgr points extracted by ( CI ) and (C?) are sometimes broad along an edge of a surface. In checking (C?), even i f some i ^ i s found at one side of the edge, (xQ,yrj) is considered as a transient point i f for some i ? at the other side,

The edge point, i s c l a s s i f i e d i n t o t h r e e c a t e g o r i e s : ( 1 ) t y p e B which s a t i s f i e s (C2) f o r b o t h i >0 and io ( ? ) t y p e L which s a t i s f i e s (C?) f o r o n l y i >0 , and ( 3 ) t y p e R w h i c h s a t i s f i e s (C?) f o r o n l y i