The development of new computational methods for use by forensic footwear examiners in the United States addresses two scenarios encountered by the forensic footwear examiner: in the investigative stage, determination of the source of an impression given a known set of outsole prints, which is useful in homicides and assaults where there are no known prints to match; and in the prosecutorial phase, determination of whether a particular impression evidence is from a known suspect's shoe with a quantification of similarity and uncertainty.
The results reported are among the first to achieve automation for matching crime-scene prints to a database of known prints. The performance is apparently significantly better than the results of the only other effort reported in the literature; however, the efficiency of the algorithms must be improved before they can be useful for the practitioner. Some of the tasks remaining include converting parts of the code from MATLAB into C++, creating additional user interfaces where user input can be solicited, and conversion of the results into a form suitable for courtroom presentation. Regarding methodology, a review was conducted of methods of footwear print examination as practiced in the United States, along with the published literature on algorithms for footwear impression analysis. Several subproblems were identified as requiring solutions: image process in order to improve the quality of the image for further automatic processing, extraction of feature for class characterization, methods for measuring the similarity of prints for the purpose of ranking the database, identifying distinctive features for individualization, and characterizing uncertainty in individualization. How these issues were addressed are described. 44 figures, 10 tables, and a 61-item bibliography