Stereovision for Real-Time Measurement of the Depth of Non-Contrastive 3-D Space by Projecting Random Patterns

  • Jun’ichi Takeno
  • Kazuyoshi Sakai
  • Shin’ichi Hatiyama
Part of the Microprocessor-Based and Intelligent Systems Engineering book series (ISCA, volume 9)


Reported here is an active type stereovision system that we developed for the real-time depth measurement of a three-dimensional space by projecting certain random patterns. The system can measure non-contrastive environments an objects in three-dimensional space. We call the system “ SVRP (Stereo Vision with Random Pattern Projection)”.

The system consists of a standard stereovision unit composed of 2 lenses and 2 CCD cameras, and a projector to project random patterns.

The internal processing of the system enables almost real-time processing using a unique method which we call “ Laminated Difference Method (LDM).”

The basic principle of this method is to determine corresponding points by comparing the lightness of each picture element in the right and left pictures as the basic information. This then allows the achieving of high-speed processing and the avoiding of occlusion problems. This method does not use the conventionally suggested method of determining corresponding points by calculating the correlation between domains of a certain size.

To reduces the vagueness of the correspondence relationship that occurs when only comparing lightness to determine corresponding points, we retrieve corresponding points using heuristic geometrical relationships among object in a space, or increase the corresponding points and further clarify correspondence by projecting ramdom patterns onto objects.

The random patterns projected are slit patterns of about 90 slit beams of random lightness for a CCD(256 x 256) plane.

Projecting random patterns offers an advantage in that the projected patterns are almost unaffected by the active patterns projected by other mobile robots.

Also, this visual sense system can be used for visual sensing by mobile robots during nighttime.

This system can process 256 horizontal scanning lines in 1.7 seconds using 10 MHz NEC PC Pascal language, that is, from the taking in of a picture to the obtaining of a measurement result of the depth in a visual field.


Mobile Robot Scanning Line Random Pattern Stereo Vision Pattern Beam 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. [1]
    Hachiyama, S., Takeno, J. A Fundamental Studies on a Visual Collision Avoidance System for Robots. Study 1.4th lecture meeting of the Robot Society of Japan, No. 2613, 1986Google Scholar
  2. [2]
    Takeno, J.: Report on Development of a Collision Avoidance Robot, Robot journal(JIRA), 1989Google Scholar
  3. [3]
    Ohta, Y. et al.: Pattern Matching Using Dynamic Programming, Information Processing Society of Japan, Vol. 30, No. 9, 1989Google Scholar
  4. [4]
    Yagi, Y. et al.: Collision Avoidance Using Ominidirectional Image Sensor(COPIS), IEEE Proc. of Int. Conf. on Robotics and Automation, P.910, 1991Google Scholar
  5. [5]
    Barnea, D. I. Silverman, H. F. A class of algorithms for fast digital image registration, IEEE Trans., C-21, 2, 1972Google Scholar
  6. [6]
    Takeno, J. Hachiyama, S.: New Technology on Stereo Vision for Mobile Robots. Proc. of 5th Int. Conf. on Advanced Robotics(ICAR), 1991Google Scholar

Copyright information

© Springer Science+Business Media Dordrecht 1991

Authors and Affiliations

  • Jun’ichi Takeno
    • 1
  • Kazuyoshi Sakai
    • 1
  • Shin’ichi Hatiyama
    • 1
  1. 1.School of Science and TechnologyMeiji UniversityTama-ku, Kawasaki-shi, Kanagawa-kenJapan

Personalised recommendations