Method and apparatus for automatic optical recognition of road signs

• US 5,633,944 A
• Filed: 04/19/1995
• Issued: 05/27/1997
• Est. Priority Date: 04/19/1994
• Status: Expired due to Fees

• Current Assignee: Automobiles Citroen (Stellantis NV), Automobiles Peugeot (Stellantis NV)
• Original Assignee: Automobiles Citroen (Stellantis NV), Automobiles Peugeot (Stellantis NV)
• Inventors: Attia, Mondher, Guibert, Laurent, Servel, Alain
• Primary Examiner(s): JOHNS, ANDREW W
• Application Number:
  US08/424,084
• Time in Patent Office: 769 Days
• Field of Search: 382/100, 382/104, 382/210, 382/211, 364/822, 359/559, 359/561, 356/351, 356/360
• US Class Current: 382/100
• CPC Class Codes: G06V 10/88 Image or video recognition using optical means, e.g. reference filters, holographic masks, frequency domain filters or spatial domain filters
• CPC Class Codes: G06V 20/582 of traffic signs
• CPC Class Codes: G08G 1/096758 where no selection takes place on the transmitted or the received information
• CPC Class Codes: G08G 1/096783 where the origin of the information is a roadside individual element

Abstract

A process and device for automatic recognition of an object such as a road sign includes a laser beam source providing a laser beam capable of passing through a modulator on which an intensity spectrum of reference signs and a sign to be recognized has been pre-recorded. A converging lens positioned downstream of the modulator, as well as devices for observing and/or analyzing the spectrum occurring at the focal plane of the lens. These devices include apparatus for detecting peaks indicating that a reference sign is identical to a sign to be recognized, as well as apparatus to separate those peaks from spurious peaks. As an example, this device can automatically recognize speed limit signs on roads or freeways.

Claims

1. A process for automatic recognition of an object comprising the steps of:

displaying on a screen an input image containing the image of the object to be recognized as well as images of reference objects;

causing a first laser beam to pass through this screen and an initial lens in order to focus the first laser beam on a modulator capable of recording an intensity spectrum;

illuminating the modulator with a second laser beam;

causing said second laser beam to pass through a second lens as the second laser beam leaves the modulator in order to produce a second spectrum supplied to a detection system; and

determining whether the image of the object to be recognized is identical to one of the images of reference objects based on the second spectrum;

wherein the step of determining involves detecting correlation peaks of the various objects contained in the input image, within the focal plane of the second lens, in such a way that the presence of intensity peaks exceeding a predetermined threshold and offset with respect to the optical center of the image indicates that the image of the object to be recognized is identical to one of the images of reference objects; and

wherein the portion of the beam corresponding respectively to said intensity peaks are polarized differently, and wherein said portions are separated in polarization in order to send them onto different detectors.

2. The process according to claim 1, wherein the input image originates as a landscape image containing the object to be recognized and undergoes a contour detection in order to adapt the input image to the input screen and decrease the spectral range.

3. The process according to claim 2 wherein the input image contains only two reference images, and only a portion of the landscape image containing the object image to be recognized is captured so that the object image of the input image is located on a line passing between the two reference images.

4. The process according to claim 1 wherein the number of reference images is greater than two, and the image to be recognized is compared, in succession, to each of several groups containing a maximum of two references images each.

5. The process according to claim 1, wherein the object to be recognized is a road sign and the images of reference objects are images of reference signs.

6. A device for automatic recognition of an image, comprising:

a modulator in which a first intensity spectrum of reference images and an image to be recognized has been pre-recorded;

a laser beam source for directing the laser beam with an optical axis so as to pass through the modulator so as to generate a second intensity spectrum having peaks and spurious peaks, the peaks indicating that one of the reference images is identical to the image to be recognized;

a converging lens placed downstream of the modulator and receiving the laser beam after the laser beam passes through the modulator;

means for analyzing the second intensity spectrum occurring at the focal plane of the converging lens, including means for detecting and separating the peaks of the second spectrum indicating that a reference image is identical to the image to be recognized from the spurious peaks; and

wherein the means for analyzing includes a mask for differentially polarizing portions of the beam containing the second intensity spectrum having the various peaks, and polarization separation optics placed downstream of the mask.

7. The device according to claim 6, wherein the mask is a plate positioned within the focal plane of the lens, having four symmetrical areas with respect to the optical axis of the beam, where two symmetrical areas with respect to the center of the plate are half-wave plates, whereas the other two symmetrical areas with respect to the center of the plate do not polarize.

8. The device according to claim 7 wherein the means for detecting includes linear CCD cameras positioned downstream of a spherical lens in combination with a cylindrical lens that, in turn, is positioned between the plate and the polarization separation optics.

9. A process for automatic recognition of a road sign, comprising the steps of:

displaying on a screen an input image containing the image of the sign to be recognized as well as images of reference signs;

causing a first laser beam to pass through this screen and an initial lens in order to focus the first laser beam on a modulator capable of recording an intensity spectrum;

illuminating the modulator with a second laser beam;

causing said second laser beam to pass through a second lens as the second laser beam leaves the modulator in order to produce a second spectrum supplied to a detection system;

determining whether the image of the sign to be recognized is identical to one of the images of reference signs based on the second spectrum;

wherein the step of determining involves detecting correlation peaks of the various signs contained in the input image, within the focal plane of the second lens, in such a way that the presence of intensity peaks exceeding a predetermined threshold and offset with respect to the optical center of the image indicates that the image of the sign to be recognized is identical to one of the images of reference signs; and

wherein the input image originates as a landscape image containing the sign to be recognized and undergoes a contour detection in order to adapt the input image to the input screen and decrease the spectral range.