System for identifying materials by NIR spectrometry

• US 5,822,219 A
• Filed: 05/13/1996
• Issued: 10/13/1998
• Est. Priority Date: 05/13/1996
• Status: Expired due to Term

• Current Assignee: Perstorp Analytical, Inc.
• Original Assignee: Foss Nirsystems, Inc. (Metrohm Stiftung)
• Inventors: Chen, Xiaolin, Monfre, Stephen L
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): KEMPER, MELANIE A
• Application Number:
  US08/645,104
• Time in Patent Office: 883 Days
• Field of Search: 364/496, 364/497, 364/498, 364/499, 364/578, 250/339.09, 250/339.11, 250/339.12, 250/340, 250/341.5, 250/341.8, 356/303, 356/306
• US Class Current: 702/27
• CPC Class Codes: G01N 21/359 using near infrared light
• CPC Class Codes: G06F 18/2135 based on approximation criteria, e.g. principal component analysis
• CPC Class Codes: G06F 18/231 Hierarchical techniques, i.e. dividing or merging pattern sets so as to obtain a dendrogram
• CPC Class Codes: G06V 10/7625 Hierarchical techniques, i.e. dividing or merging patterns to obtain a tree-like representation; Dendograms
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G16C 20/20 Identification of molecular entities, parts thereof or of chemical compositions

Abstract

In a method for identifying an unknown product a library of absorbance spectra of known products is measured and stored in a library. A quick search using clustering techniques is conducted to narrow the search to a few products, followed by an exhaustive search of the spectra of the few products. More specifically, principal component analysis is applied to the absorbance spectra to generate product score vectors extending into principal component inside model space which are divided into clusters and subclusters in accordance with their relative proximity. Hyperspheres are constructed around each vector and an envelope is constructed to enclose each cluster surrounding the hyperspheres within the cluster. The absorbance spectrum of the unknown product to be identified is measured and an unknown product score vector is determined from the unknown product spectrum projecting in principal component inside model space of the clusters. It is determined whether or not the unknown product score vector falls within one of the envelopes and if so the product score vector is projected into the principal component inside model space of that cluster and it is determined whether or not the unknown product score vector falls within any of the subclusters divided from the cluster. This process is repeated until the unknown product score vector is found to lie in a cluster which is not further subdivided. In this manner, the search is narrowed to a few products. An exhaustive search is then carried out to match the spectrum of the unknown product with the spectra of the known products corresponding to the undivided subcluster.

Claims

1. A method of matching an unknown product with one of a library of known products comprising the following steps:

step (1);

measuring a near infrared absorbance spectrum for each of said known products,

step (2);

generating known product vectors extending into hyperspace representing the absorbance spectra determined for each of said known products,

step (3);

dividing said known product vectors into clusters of vectors extending into hyperspace wherein the end points of the vectors of each cluster are closer to each other in hyperspace than the vectors outside of such cluster,

step (4);

dividing at least some of said clusters of vectors into subclusters of vectors extending into hyperspace,

step (5);

repeating said step (4) on at least some of said subclusters until all of said subclusters have fewer than a predetermined number of vectors,

step (6);

surrounding the end points of the vectors of each of said clusters and subclusters with an envelope defined in the corresponding hyperspace,

step (7);

measuring the absorption spectrum of said unknown product,

step (8);

determining in which of said envelopes surrounding said clusters divided in step (3) the end points of a vector, representing said unknown product and extending into the hyperspace of said clusters, falls,

step (9);

if the end point of a vector representing said unknown product falls into an envelope surrounding a cluster which is divided into subclusters, then determining in which envelope surrounding a subcluster the end point of a vector representing said unknown product and extending into the hyperspace of such subcluster, falls,

step (10);

repeating the step (9) on further divided subclusters until a vector representing said unknown product is determined to fall into an envelope surrounding a subcluster which is not further divided,

step (11);

then determining which known product represented by a vector within said last-named envelope said unknown product matches.

2. A method as recited in claim 1 wherein said step (2) includes subjecting said absorbance spectra determined for said known products to principal component analysis to determined a known product score vector representing each known product projected into principal component inside model space.

3. A method as recited in claim 2 wherein said step (4) includes subjecting the absorbance spectra of the known product vectors of each cluster and subcluster to principal component analysis to determined a product score vector for each known product of the cluster or subcluster extending into principal component inside model space determined for the known product spectra of such cluster or subcluster.

4. A method as recited in claim 2 further comprising surrounding the end point of each known product score vector with a hypersphere, said step (6) including surrounding the hyperspheres of the corresponding clusters with each of said envelopes.

5. A method as recited in claim 2 wherein each of said envelopes comprises a hyperbox of orthogonal dimensions.

6. A method as recited in claim 1 further comprising surrounding the end point of each of said known product vectors with a hypersphere, said step (6) including surrounding the hyperspheres of the corresponding clusters with said envelopes.

7. A method as recited in claim 1 wherein each of said envelopes comprises a hyperbox having orthogonal dimensions.

8. A method of matching a product with one of a library of known products comprising

measuring the absorbance spectra of said known products,

subjecting said absorbance spectra to principal component analysis to determine known product score vectors projecting in principal component inside model space,

surrounding the end points of said known product score vectors with hyperspheres,

measuring the absorbance spectrum of said unknown product,

determining from the spectrum of said unknown product an unknown product score vector projecting in principal component inside model space, and

determining in which of said hyperspheres said unknown product score vector falls to thereby determine which known product said unknown product matches.

9. A method as recited in claim 8 wherein said absorbance spectra of said known products includes a training set of a plurality of spectra for each of said known products and wherein said method includes obtaining an average absorbance spectra of each training set, determining from said average spectra and the corresponding training set a standard deviation for each known product, the hyperspheres each having a radius equal to a selected multiple of said standard deviation.