Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The CSL vision system

The CSL vision system THE GSL VISION SYSTEM by Wesley E. Snyder Coordinated Science Laboratory University of Illinois Urbana Illinois Thls is a summary of a paper by the same title publlSbed aS Al Tecn ~ote 5, Coordinated Science Laboratory. The longer report contains detailed results of the tests described herein. 6a The CSL vision system is the result of an attempt to combine the Pest features of existing vision electronic~ into a computer vision system. This report briefly describes some of the more interesting components of the system, performance tests, anO o p e r a t i o n a l evaluation. The heart of the system is a silicon vidlcon. Silicon vidicuns differ from c o n v e n t i o n a l vldicons in that the imaging device is a silicon wafer with an array of didoes diffused into Its surface. When ll~nt hits these diodes, excess charges are created which are swept out by a scanning electron beam, producing a current whicn is p r o p o r t i o n a l to the light intensity. Advantages of silicon vidicons include a very wide spectral response ana an ability to be virtually burnout-proof, our particular vidicon also has a special i o n - i m p l a n t e d layer to make it more resistant to blooming (apparent e n l a r g e m e n t of intensely llghted regions). We employ a c o n v e n t i o n a l television camera, Modified to make use of the special vidicon. One interesting problem did arise, however. The output of c o n v e n t i o n a l TV cameras is AC coupled. That is, the averaKe value of the output voltage is a ground potential. TOe effect of that is that the voltaKe c o r r e s p o n d l n g to "blacK" differs from one picture to the next. Since the a n a l o g - t o - a i g i t a l converter converts that voltage to a number representing intensity, the number c o r r e s p o n d i n g to %ne same intensity varies from scene to scene. This problem l~as been solved in a number cf ways, including using software to normalize the picture, and using a s a m p l e - a n d - b o l d circuit to reference each Point tO a voltage known to represent blacK. We solved it ty building a "video clamp" which clamps the lowest voltage (black) to ground potential. The a n a l o g - t o - d i g i t a l converter we are using makes use of the latest techniques in the design o~ low-distortion, high speed converters. It gives an eight bit sample every 200ns. The system is interfaced to a PDP-iO oy a high-speed i n t e r f a c e WhiCh reads an entire picture Into memory in 16ms. The picture consists of an array of 238 lines by 252 samples per line. ~ach sample is six bits, although the programmer can set the interface to select the low, middle, or upper six of the eiKnt bits fro~, the A/D converter. No sampling occurs during cai~era blanking, so all R>tx252 points are usable picture data. We set up the system to ru~ under full timesnaring, but during the 16 ms that the interface is actually transferring data to the computer, %f any other device |including ~he CPU) attempts tc access memory, lost data will result. Insuring that memory was unused ~nile maintaining timesharing provided an interesting programming problem. Performance of the System ou Noise: Noise could be defined in terms of the Fourier spectrum oz the output of the camera; however, we tried to define noise in an o p e r a t i o n a l manner, more a~in to now the noise actually affects the users of a vision system: a picture was taken of a sheet of photographic paper having homogeneous reflectivity. Each point in tne picture was examlned and compared to ~ts o-neighbors. If the intensity of the point was greater than toe maximum of its neignbors, or less than their minimum, it was considered to be a noise point. Several pictures were taken at two different background illuminations. On the average, about 2~0 points were noise, out oz 60000. They were distributed in intensity aoout the ambient intensity in a Gaussian manner. Blooming: The camera was focussed on a circle of light, under n o n - o v e r l o a d conditions, then the intensity of the spot was varied over orders of magnitude, with pictures being taken and the spot size measured. The system was originally specified to have oloomlng c h a r a c t e r i s t i c s Such that "a spot of size originally 1% of image diagonal shall not bloom to more than 5% at i000000 tlmes overload. ~ our m e a s u r e m e n t s indicate that the system does not meet those lofty standards, put it does exceed the best specifications of any other vidicon, silicon or otherwise. Distortion! The only measurable distortion is "parabolic distortion" which, due to electron optical effects, causes the corners of the picture to seem darker than the center. This can be corrected by a piece of hardware, called a "parabolic correction amplifier," a step we ~ave not taken as Yet. LinearitY: Within the operation intensity. range of the camera, tlle output voltage varies exactly linearily witn the There ~re still m a n y unanswered questions concerning tne proper way to characterize the performance of a vision system. For example, what is tl~e best way to c h a r a c t e r i z e noise? How important is it that the response to light be linear? I would be interested in c o r r e s p o n d i n g with others who have had exoerience with vision hardware and would like to try to establAsn Some uniform parameters With which tO evaluate vision systems. 6c April 197h SIGA~T NEWSLETTER ~age 7 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ACM SIGART Bulletin Association for Computing Machinery

The CSL vision system

ACM SIGART Bulletin , Volume (45) – Apr 1, 1974

Loading next page...
 
/lp/association-for-computing-machinery/the-csl-vision-system-NubiAAn63F

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Association for Computing Machinery
Copyright
Copyright © 1974 by ACM Inc.
ISSN
0163-5719
DOI
10.1145/1045220.1045221
Publisher site
See Article on Publisher Site

Abstract

THE GSL VISION SYSTEM by Wesley E. Snyder Coordinated Science Laboratory University of Illinois Urbana Illinois Thls is a summary of a paper by the same title publlSbed aS Al Tecn ~ote 5, Coordinated Science Laboratory. The longer report contains detailed results of the tests described herein. 6a The CSL vision system is the result of an attempt to combine the Pest features of existing vision electronic~ into a computer vision system. This report briefly describes some of the more interesting components of the system, performance tests, anO o p e r a t i o n a l evaluation. The heart of the system is a silicon vidlcon. Silicon vidicuns differ from c o n v e n t i o n a l vldicons in that the imaging device is a silicon wafer with an array of didoes diffused into Its surface. When ll~nt hits these diodes, excess charges are created which are swept out by a scanning electron beam, producing a current whicn is p r o p o r t i o n a l to the light intensity. Advantages of silicon vidicons include a very wide spectral response ana an ability to be virtually burnout-proof, our particular vidicon also has a special i o n - i m p l a n t e d layer to make it more resistant to blooming (apparent e n l a r g e m e n t of intensely llghted regions). We employ a c o n v e n t i o n a l television camera, Modified to make use of the special vidicon. One interesting problem did arise, however. The output of c o n v e n t i o n a l TV cameras is AC coupled. That is, the averaKe value of the output voltage is a ground potential. TOe effect of that is that the voltaKe c o r r e s p o n d l n g to "blacK" differs from one picture to the next. Since the a n a l o g - t o - a i g i t a l converter converts that voltage to a number representing intensity, the number c o r r e s p o n d i n g to %ne same intensity varies from scene to scene. This problem l~as been solved in a number cf ways, including using software to normalize the picture, and using a s a m p l e - a n d - b o l d circuit to reference each Point tO a voltage known to represent blacK. We solved it ty building a "video clamp" which clamps the lowest voltage (black) to ground potential. The a n a l o g - t o - d i g i t a l converter we are using makes use of the latest techniques in the design o~ low-distortion, high speed converters. It gives an eight bit sample every 200ns. The system is interfaced to a PDP-iO oy a high-speed i n t e r f a c e WhiCh reads an entire picture Into memory in 16ms. The picture consists of an array of 238 lines by 252 samples per line. ~ach sample is six bits, although the programmer can set the interface to select the low, middle, or upper six of the eiKnt bits fro~, the A/D converter. No sampling occurs during cai~era blanking, so all R>tx252 points are usable picture data. We set up the system to ru~ under full timesnaring, but during the 16 ms that the interface is actually transferring data to the computer, %f any other device |including ~he CPU) attempts tc access memory, lost data will result. Insuring that memory was unused ~nile maintaining timesharing provided an interesting programming problem. Performance of the System ou Noise: Noise could be defined in terms of the Fourier spectrum oz the output of the camera; however, we tried to define noise in an o p e r a t i o n a l manner, more a~in to now the noise actually affects the users of a vision system: a picture was taken of a sheet of photographic paper having homogeneous reflectivity. Each point in tne picture was examlned and compared to ~ts o-neighbors. If the intensity of the point was greater than toe maximum of its neignbors, or less than their minimum, it was considered to be a noise point. Several pictures were taken at two different background illuminations. On the average, about 2~0 points were noise, out oz 60000. They were distributed in intensity aoout the ambient intensity in a Gaussian manner. Blooming: The camera was focussed on a circle of light, under n o n - o v e r l o a d conditions, then the intensity of the spot was varied over orders of magnitude, with pictures being taken and the spot size measured. The system was originally specified to have oloomlng c h a r a c t e r i s t i c s Such that "a spot of size originally 1% of image diagonal shall not bloom to more than 5% at i000000 tlmes overload. ~ our m e a s u r e m e n t s indicate that the system does not meet those lofty standards, put it does exceed the best specifications of any other vidicon, silicon or otherwise. Distortion! The only measurable distortion is "parabolic distortion" which, due to electron optical effects, causes the corners of the picture to seem darker than the center. This can be corrected by a piece of hardware, called a "parabolic correction amplifier," a step we ~ave not taken as Yet. LinearitY: Within the operation intensity. range of the camera, tlle output voltage varies exactly linearily witn the There ~re still m a n y unanswered questions concerning tne proper way to characterize the performance of a vision system. For example, what is tl~e best way to c h a r a c t e r i z e noise? How important is it that the response to light be linear? I would be interested in c o r r e s p o n d i n g with others who have had exoerience with vision hardware and would like to try to establAsn Some uniform parameters With which tO evaluate vision systems. 6c April 197h SIGA~T NEWSLETTER ~age 7

Journal

ACM SIGART BulletinAssociation for Computing Machinery

Published: Apr 1, 1974

There are no references for this article.