MUG FILE PROJECT REPORT NUMBER UHMUG-11 - FORGERY APPLICATION OF A PATTERN RECOGNITION ALGORITHM FOR FACIAL IMAGES

• NCJ Number: 47590
• Author(s): B T RHODES; K PRASERTCHUANG
• Date Published: 1977
• Length: 42 pages
• Grant Number(s): 76-NI-99-012
• Sponsoring Agency: US Dept of Justice
• Corporate Author: University of Houston
• Sale Source: National Institute of Justice/
• Publication Type: Report (Study/Research)
• Language: English
• Country: United States of America

Annotation

THIS REPORT CONSIDERS THE APPLICATIONS OF A PATTERN RECOGNITION ALGORITHM TO THE PROBLEM OF MATCHING PHOTOGRAPHS OF THE SAME INDIVIDUAL IN A SET OF PHOTOGRAPHS AVAILABLE TO FORGERY INVESTIGATORS.

Abstract

THE PROBLEM CONSIDERED IN THIS REPORT ARISES IN FORGERY CASES AND IN OTHER IDENTIFICATION SITUATIONS THAT DEPEND ON PHOTOGRAPHS OF FACES. IN FORGERY CASES, A PHOTOGRAPH OF THE PERSON WHO CASHED A FORGED CHECK IS OFTEN AVAILABLE. OVER A PERIOD OF TIME, THE FORGERY UNIT OF A LARGE POLICE DEPARTMENT WILL ACCUMULATE A SIZABLE SET OF THESE PHOTOGRAPHS AND MAY RECEIVE 15-50 NEW PHOTOGRAPHS EACH WEEK. THE OBJECTIVE OF THIS RESEARCH WAS THE DEVELOPMENT OF A COMPUTER-AIDED SYSTEM WHICH COULD MATCH THE INDIVIDUALS SHOWN IN NEW PHOTOGRAPHS COMING INTO A FORGERY DEPARTMENT WITH OTHER PHOTOGRAPHS OF THE SAME INDIVIDUAL ALREADY IN THE DEPARTMENT'S FILE. THE APPROACH WAS BASED ON A MUG SHOT RETRIEVAL SYSTEM WHICH USED A PATTERN RECOGNITION ALGORITHM TO DETERMINE THE SIMILARITY OF FACIAL IMAGES IN THE STANDARD MUG SHOT POSITION. THERE ARE TWO FUNDAMENTAL DIFFERENCES IN THE FORGERY APPLICATION: (1) A PHOTOGRAPH IS ALREADY AVAILABLE, SO THE DEPARTMENT DOES NOT HAVE TO RELY ON THE MEMORY OF A WITNESS; AND (2) THE PHOTOGRAPHS IN THIS APPLICATION MAY BE TAKEN WITH THE CAMERA IN A DIFFERENT POSITION FROM THE STANDARD MUG SHOT POSITION. THE MAJOR RESEARCH PROBLEM WAS THE DEVELOPMENT OF A 'ROTATE' PROCEDURE TO HANDLE PHOTOGRAPHS TAKEN FROM A NONSTANDARD POSITION. THE DEVELOPMENT OF THE ROTATE PROCEDURE IS DISCUSSED. THE PROCEDURE INVOLVES TAKING SPECIFIED MEASUREMENTS FROM THE NONSTANDARD PHOTOGRAPH AND CALCULATING AN ESTIMATE OF THE VALUES WHICH WOULD HAVE BEEN OBTAINED IF THE PHOTOGRAPH HAD BEEN TAKEN IN THE FORWARD-BUST VIEW USED IN THE STANDARD MUG SHOT. THE EXPERIMENTAL AND FIELD TRIAL RESULTS ARE EXAMINED. AN ANALYSIS OF 258 PHOTOGRAPHS FROM POLICE FORGERY FILES INDICATED THAT ABOUT 50 PERCENT MET STANDARD POSITION CRITERIA AND THAT AN ADDITIONAL 40 PERCENT FELL WITHIN THE APPLICABLE LIMITS OF THE ROTATE PROCEDURE. THE FUTURE POSSIBILITIES OF THE PROCEDURE ARE ALSO DISCUSSED. APPENDED MATERIALS INCLUDE TRACINGS OF STANDARD AND NONSTANDARD PHOTOGRAPHS USED IN THE EXPERIMENT AND THE PROPOSED GUIDELINES FOR FIELD TRIALS. FOR RELATED DOCUMENTS, SEE NCJ 47585-47589 AND 47591-47592. (AUTHOR ABSTRACT MODIFIED--JAP)