To help the USPS to protect its source of revenue, SRI has developed algorithms that identify
mail pieces with no postage or too little postage. Currently the algorithms identify mail pieces
with one type of facing identification mark (FIM) but without any postage, and mail pieces with only
a single one- or three-cent stamp. These algorithms are based on unique SRI techniques for finding
stamps in mail-piece images and reading the denominations of the stamps, regardless of their
orientation, complex and colorful backgrounds, and low-resolution text (Figure 1).
SRI's revenue protection technology has a false-positive rate of 15% and a false-negative rate of 8%. SRI's
revenue protection technology is integrated into the video facing system that is being deployed
nationwide. The algorithms are currently being extended to identify mail pieces with only a single
two- or five-cent stamp, and will ultimately be extended to identify mail pieces with any amount of
postage that is less than the minimum.
Figure 1. Extraction of Stamp Denomination Text