Machine vision for adaptive laser beam steering

• US 5,576,948 A
• Filed: 10/06/1994
• Issued: 11/19/1996
• Est. Priority Date: 07/28/1992
• Status: Expired due to Fees

• Current Assignee: Rudolph Technologies Incorporated (Onto Innovation Inc.)
• Original Assignee: Robotic Vision Systems, Inc. (Omron Corporation)
• Inventors: Elstein, Stanley, Stern, Howard, Maali, Fereydoun
• Primary Examiner(s): Trammell, James P.
• Application Number:
  US08/319,415
• Time in Patent Office: 775 Days
• Field of Search: 364/167.01, 364/559-565, 364/474.03, 364/474.05, 364/474.37, 382/103, 382/106, 382/107, 382/145, 382/146, 382/153, 382/154, 382/173-180, 382/171, 356/4.01, 356/3.01, 356/3.03, 356/3.09, 356/3.06, 356/3.07
• US Class Current: 700/59
• CPC Class Codes: G05B 19/408 characterised by data handling or data format, e.g. reading, buffering or conversion of data
• CPC Class Codes: G05B 2219/37053 Optical triangulation
• CPC Class Codes: G05B 2219/37553 Two cameras one for coarse scanning, other for fine scanning
• CPC Class Codes: G05B 2219/37571 Camera detecting reflected light from laser
• CPC Class Codes: G05B 2219/45034 Adjusting, trimming circuits on printed boards

Abstract

An arrangement for steering the laser beam, in an active optical triangulation method based on three-dimensional sensors, along paths which are sensitive to the context of the scene of interest, and alleviating the need to undergo a search path by the sensor. A conventional video camera is used to preview the scene of interest and the resultant intensity image is processed to locate the desired path based on the scene context. The sensory system is modeled to enable determination of the instantaneous angular deflections of the laser beam along with the translation of the laser source that can shift the nominal or actual data acquisition path to the desired data acquisition path.

Claims

1. A method for steering a laser beam of an active optical triangulation-based three-dimensional sensor by information extracted from an intensity image of a scene previewed by a video camera through laser beam deflection, comprising the steps of:

generating an equation of a nominal laser spot path, at a datum plane;

generating an equation of a desired path;

generating an equation of a laser plane;

generating coordinates of a laser point source;

projecting all said coordinates onto a global three-space coordinate system;

locating intersections of said spot path and said desired path with the said laser plane;

computing a subtended angle of said intersections on said laser plane when viewed from a laser source point; and

deflecting the laser beam through said computed subtended angle.

2. A method as defined in claim 1, wherein the step of generating an equation of a laser plane comprises projecting the laser beam at substantially two different angles and recording coordinates of illuminated points on a surface, of a calibration block while being raised or depressed through a view volume of the said three-dimensional sensor;

fitting said recorded coordinates of said illuminated points in a local space of the said three-dimensional sensor to a resultant plane by a least squares method; and

designating a laser plane in the location of said fitted plane.

3. A method as defined in claim 2, including the additional steps of:

fitting elements of each point set derived from each of said different angles to a line using the method of least squares;

projecting thereafter said fitted lines onto the equation of said laser plane, forming two coplanar lines intersecting at a point; and

designating the coordinates of the laser point source to be at the location of said intersecting point.

4. A method as defined in claim 1, wherein said information extracted from an intensity image utilizes adaptive thresholding of a gray level image to binarize the image comprising the steps of:

acquiring the image and constructing its original gray level histogram;

area quantizing said histogram;

generating a discrete function from a span of successive area quanta;

differentiating said function and locating two successive maxima of said function;

locating

5. A method as defined in claim 1, including the steps of:

pivoting said video camera into positions ion either side of said three-dimensional sensor for previewing a scene of interest with respect to said three-dimensional sensor.

6. A method as defined inn claim 1, including the steps of:

providing a lens assembly detached from said video camera;

translating said lens assembly to positions on either side of said three-dimensional sensor for imaging a scene of interest onto a retina of said video camera in front of said three-dimensional sensor.

7. A method for translating a laser point source in an active optical triangulation-based three-dimensional sensor guided by information extracted from an intensity image of a scene previewed by a two-dimensional sensor through three-dimensional sensor translation, comprising the steps of:

generating an equation of a nominal laser spot path at a datum plane;

generating an equation of a desired path;

projecting an equation of said spot path and said desired path onto said datum plane;

intersecting said spot path and said desired path at a point of intersection and at an angle between the paths;

deriving coordinates of said intersection point and said angle between the paths;

computing sensor translation at any point on the said projected nominal laser spot path and distance from said point of intersection, said translation being the product of said distance and tangent of said angle; and

translating the three-dimensional sensor through said computed sensor translation.

8. A method for segmentation of lead toe points and subsequent extraction of a lead toe line from a binarized image of a quad flat pack device comprising the steps of:

forming a lateral projection vector of said binarized image;

finding a maximum element value in the said projection vector and clipping all elements of the said vector at a predetermined fraction of the said maximum;

run-length coding said clipped projection vector so that each run signifies a sequence of non-zero elements;

generating an objective function for each run based on a product of each run-length by a weighing function representing cumulative sum of the points supported by that run of the projection vector;

locating the run exhibiting said maximum objective function and proceeding along ascending or descending row numbers, dependent on orientation of the device, to locate the next run number, and if the next run number is non-existent the run with maximum objective function is retained instead;

locating a maximum value of said projection vector within bounds of the retained run number, and arriving thereby at a datum row number for subsequent formation of a vertical projection vector;

forming a vertical projection vector for a part of an image that extends to either side of the datum row number by a predetermined extent, elements of said vertical projection vector being run-length encoded so that each run signifies a consecutive sequence of non-zero elements;

computing an objective function for each run as a product of run length and a weighing factor derived from cumulative sum of the projection vector within bounds of the run;

finding a mode of the objective functions and suppressing runs exhibiting objective functions deviating by more than a predetermined value from said mode;

initiating a search at columns passing through a center of surviving runs in an ascending or descending row number direction dependent on orientation of the device for sustained transition from 1 to 0 in the binarized image for obtaining a resulting point set corresponding to lead toe points subjected to fast Hough transform yielding a longest subset of colinear points supporting the lead toe line;

fitting said colinear points to a line using a least square method; and

designating a lead toe line in the location of said fitted line.