Unisys Corporation

Other

Craig Weaver Harris

Kha Nguyen

Theodore C White

Jayesh V Sheth

Chung W Wong

Richard A Cowgill

James Downie

Ulrike Helene Rinehart

Eddie Louis Madison

Alan Thomas Theis

Julian Garrard Blackburn

George Moffat

A method of recognizing an unknown object includes the following steps: receiving from a camera, a first image of the unknown object which is comprised of an array of pixels of multiple magnitudes that lie within a certain range; generating a second image from the first image by scaling the pixel magnitudes in the first image such that differences in the scaled pixel magnitudes are increased and the average pixel magnitude lies near the middle of the pixel range; and, determining the width of the unknown object by detecting an area in the second image where the pixel magnitudes exceed a certain threshold level. Based on that width, a third image is generated from the second image, by adjusting the magnification of the second image such that it has a predetermined standard width; and the unknown object is identified if a correlation between the third image and a reference exceeds a threshold.

Method of compensating for pixel histogram distortion in an object recognition system

An input image is comprised of an array of pixels in rows and columns having P.sub.R pixels-per-row and P.sub.C pixels-per-column. To generate a scaled replica of the input image, the number of pixels-per-row P.sub.R ' and the number of pixels-per-column P.sub.C ' that are to be in the scaled replica are selected, where P.sub.R ' and P.sub.C ' are positive integers which can be larger or smaller than P.sub.R and P.sub.C. Then the input image is distorted such that the spacing between columns is (P.sub.R '-1)x and the spacing between rows is (P.sub.C '-1)y, where x and y are arbitrary units of distance, such as one inch or one meter. Then the distorted input image is overlaid with a new row-column array, where the spacing between columns is (P.sub.R -1)x and the spacing between rows is (P.sub.C -1)y, and where x and y are the same units of distance that are in the distorted input image. Then, using the above described spacings between the rows and the columns in the distorted input image and the overlaid new array, the pixels for the scaled replica are generated as a function of the various areas which those rows and columns delineate.

Method of generating a scaled replica of a digital input image by

A system for recognizing objects is comprised of: 1) a camera that has a lens which is uncovered; 2) a housing which encloses the camera, and has a window for the lens; 3) a rotating means, which is inside of the housing and which carries the camera, for receiving electronic control signals and rotating the camera at different angles in response thereto; and, 4) a control means for generating the electronic control signals such that the rotating means selectively points the camera lens -a) through the window at any one of several predetermined angles, and b) inside of the housing at the surface for calibrating the camera. With this system, the objects that are recognized can be human faces at different heights relative to the window in the housing. By pointing the camera lens inside of the housing, the lens is protected from damage even though it is not covered; and a calibration surface inside of the housing enables the camera to be calibrated while the lens is protected.

Apparatus for positioning a camera in an object recognition system

An input image is comprised of an array of pixels in rows and columns. To generate a scaled replica of the input image, that input image is first partitioned into several identical sections, in which each section has P.sub.R pixels per row and P.sub.C pixels per column. Then the number of pixels-per-row P.sub.R ' and the number of pixels-per-column P.sub.C ' that are to be in each corresponding section of the scaled replica is selected, where P.sub.R ' and P.sub.C ' are any positive integers which can be larger or smaller than P.sub.R and P.sub.C. Then just a single section of the input image is distorted such that the spacing between columns is (P.sub.R '-1)x and the spacing between rows is (P.sub.C '-1)y, where x and y are arbitrary units of distance, such as one inch or one meter. Then the single distorted section of the input image is overlaid with a section of a new row-column array, where the spacing between columns is (P.sub.R -1)x and the spacing between rows is (P.sub.C -1)y, and where x and y are the same units of distance that are in the distorted input image. Then, using the above described spacings between the rows and the columns in the single distorted section of the input image and the overlaid new array, the pixels for all sections of the scaled replica are generated as a function of the various areas which those rows and columns delineate.

An interbus interface module enables storage and transfer of commands, messages and data between parallel a dual system bus operating on a first protocol and a subrequestor bus operating on a second protocol. The interface module serves a first group of requestors, such as multiple processors and main memory, for handling data transfers to and from the subrequestor bus via said dual system buses while also handling data transfers to and from a second group of requestors connected to the subrequestor bus.

Interbus interface module

An Input/Output Module (IOM) interfacing multiple computers attached to a dual system bus. The IOM provides an interbus module which interfaces the dual system bus to a sub-requestor bus connecting multiple sub-requestor modules. The sub-requestor modules control a plurality of interface adaptors permitting data transfers to/from a variety of peripherals using different data protocols and clock rates. The requirements for the main host processors and memories in a computer system would be unduly burdensome were it not for the relief from these overhead operations by the input/output module which provides the tailoring of data transfer capability to and from a multiplicity of peripherals having many different types of protocols and clock rates.

Growing community of inventors

Lake Forest, CA, United States of America

Craig Weaver Harris