This article reports on a project in which search prefilters for the prototype infrared (IR) spectral library of paint data query (PDQ) database were developed from the clear coat, surfacer-primer, and e-coat layers for 1,179 manufacturer paint systems within a limited production year range (2000-2006) to identify vehicle manufacturer (Ford, Chrysler, and General Motors).
For each make (i.e., manufacturer), search prefilters were developed to identify the assembly plant of the vehicle using a hierarchical classification scheme. A cross correlation library search algorithm that performed both forward and backward searching was then used to identify the line and model of the vehicle from the truncated IR spectral library of PDQ identified by the search prefilters. Samples assigned to the same line and model by both a forward and backward search of the IR spectral data were always correctly matched, always correlated well on an individual basis to a specific library sample, and were well represented in the truncated PDQ spectral library identified by the search prefilters. The performance of the prototype IR library searching system (search prefilters and cross-correlation library search algorithms) for the PDQ database was benchmarked against commercial library searching algorithms. Only the results for Ford are reported in this article. (Publisher abstract modified)