NCJ Number
47591
Date Published
1977
Length
35 pages
Annotation
THIS REPORT DESCRIBES A SET OF SUBROUTINES WHICH ARE A FORTRAN IV CODING OF A PATTERN RECOGNITION ALGORITHM DESIGNED TO RETRIEVE LOOK-ALIKES FROM A COMPUTERIZED MUG FILE.
Abstract
THE ALGORITHM PERMITS PATTERN RECOGNITION OF FACIAL IMAGES BASED ON STANDARDIZED MEASUREMENTS ON MUG SHOT PHOTOGRAPHS. THE SUBROUTINES PRESENTED PERMIT THE ALGORITHM'S INCORPORATION INTO THE SORTING CAPABILITIES OF PROGRAMS SIMILAR TO THE OAKLAND, CALIFORNIA, POLICE DEPARTMENT'S CRIME SYSTEM. THE ALGORITHM REQUIRES NINE FACIAL MEASUREMENTS AS PRIMARY INPUT. THE MEASUREMENTS CAN BE TAKEN FROM A SKETCH OR PHOTOGRAPH OF THE SUBJECT. EACH STEP OF THE ALGORITHM PERFORMS A TRANSFORMATION ON THESE MEASUREMENTS OR THEIR RATIOS. IN THE FIRST STEP, THESE MEASUREMENTS ARE PAIRED TO FORM EIGHT RATIOS. THE SECOND STEP MODIFIES THE RATIOS GENERATED FROM A SKETCH OR COMPOSITE, BUT NOT THOSE FROM A PHOTOGRAPH. IN THE FINAL STEP, THE EUCLIDEAN DISTANCE BETWEEN THE SELECTED RATIOS OF THE IMAGE AND THE MUG SHOT ON FILE IS CALCULATED AND SORTED IN ASCENDING ORDER REGARDING SIMILARITIES BETWEEN THE SUSPECT AND THE MUG FILE PHOTOGRAPH. THE SIX FORTRAN SUBROUTINES USED IN THIS PROCEDURE ARE NOTED AND DISCUSSED. ILLUSTRATIONS OF FACIAL IMAGE MEASUREMENTS, MEASUREMENT TABLES, SKETCH AND COMPOSITE RATIOS, AND PARAMETER VALUES ARE INCLUDED FOR RELATED AND DOCUMENTS, SEE NCJ 47585-47590 AND 47592. (JAP)