System and method for taxonomically distinguishing sample data captured from biota sources

• US 9,886,945 B1
• Filed: 03/18/2014
• Issued: 02/06/2018
• Est. Priority Date: 07/03/2011
• Status: Active Grant

• Current Assignee: Reality Analytics, Inc. (Renesas Electronics Corporation)
• Original Assignee: Reality Analytics, Inc. (Renesas Electronics Corporation)
• Inventors: Sieracki, Jeffrey M.
• Primary Examiner(s): Leland, III, Edwin S
• Application Number:
  US14/218,650
• Time in Patent Office: 1,421 Days
• Field of Search: 704237
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 18/2136 based on sparsity criteria, e.g. with an overcomplete basis
• CPC Class Codes: G06V 10/443 by matching or filtering
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 17/06 Decision making techniques; Pattern matching strategies
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices

Abstract

A system and method are provided for distinguishing biota, such as insect types, from overall images, images of their wings or of other body parts. The system and method exploit various techniques described herein, in combination with large scale capture of sample imagery to achieve a flexible mechanism for automated classification on biota of any type.

Claims

1. A system for taxonomically distinguishing anatomic samples captured in unconstrained manner for a plurality of source organisms, comprising:

at least one field acquisition portion for acquiring anatomic samples of the source organisms;

a portable imagery portion disposed at said field acquisition portion, said portable imagery portion actuating to capture images of the anatomic samples acquired at said field acquisition portion;

a classifier portion remotely disposed from said field acquisition portion, and receiving the captured images as anatomic image data segments, said classifier portion actuating a server to execute;

a vector operation constructing for each of the anatomic image data segments at least one vector defined by pixel values for a predetermined strip of pixels across a corresponding one of the captured images, at least one of the anatomic image data segments being selected from a plurality of different generated versions of a captured image preserved for selection exclusive of one or more other generated versions;

a sparse decomposition operation selectively executing in at least a training system mode a simultaneous sparse approximation upon a joint corpus of pixel value vectors for a plurality of anatomic image data segments of distinct source organisms, said sparse decomposition operation adaptively generating at least one sparse decomposition for each said vector with respect to a representative set of decomposition atoms;

a discriminant reduction operation executable during the training system mode to derive an optimal combination of atoms from said representative set for cooperatively distinguishing image signals attributable to the anatomic samples of different ones of the distinct source organisms; and

,

a classification operation executable in a classification system mode to discover for said sparse decomposition of an input anatomic image data segment a degree of correlation in the optimal combination of atoms relative to the anatomic samples of each of the distinct source organisms,

different source organisms being thereby distinguished from one another based on anatomic data captured without constraint on the nature and form of data content.

2. The system as recited in claim 1, further comprising a database portion pre-stored with a plurality of anatomic data segments attributable respectively to a plurality of distinct source organisms, said classifier portion in the training system mode reading the pre-stored anatomic data segments from said database portion, said sparse decomposition and discriminant reduction operations of said classifier portion thereby executing in the training system mode to identify a distinct class corresponding to at least one of the known distinct source organisms.

3. The system as recited in claim 1, further comprising a personal communication device coupled to said portable imagery portion, said personal communication device actuating transmission of the captured images as anatomic data segments to said classifier portion.

4. The system as recited in claim 3, wherein said classifier portion actuates a responsive transmission back to the personal communication device originating an anatomic data segment transmission, the responsive transmission indicating a distinct source organism correlation discovered for the transmitted anatomic data segment.

5. The system as recited in claim 4, wherein said classifier portion filters the distinct source organism correlation discovered for the anatomic data segment transmission according to at least one of:

pre-stored data on the distinct source organism and geographic location data for the field acquisition portion.

6. The system as recited in claim 3, wherein at least one of the other portions of the system is controllable through said personal communication device implementing an online application for identifying and classifying the distinct source organism of the anatomic sample.

7. The system as recited in claim 1, wherein said portable imagery portion includes a manipulator base supporting at least one of an anatomic sample and an imaging device for angularly displacement of one relative to the other, said portable imagery portion thereby capturing the images of the anatomic sample in angularly adjustable manner.

8. The system as recited in claim 1, comprising a plurality of said field acquisition portions remotely disposed from one another, said field acquisition portions including facilities for collecting the anatomic samples from at least one of:

vehicle grills, vehicle windshields, cargo articles, freight containers, naturally growing plant materials, stored specimen depositories, specimen collection traps, clothing of travelers, and luggage of travelers.

9. The system as recited in claim 1, wherein:

the anatomic samples include at least one of;

an insect wing image type and an arachnid profile image type; and

,

said vector operation of said classifier portion includes selective down-sampling of the anatomic sample images in resolution to construct the vectors of predetermined form for the anatomic data segments.

10. The system as recited in claim 1, wherein said discriminant reduction operation of said classifier portion includes a Support Vector Machine (SVM) programmably implemented therein, said SVM executing to mutually n-wise compare the distinct source organisms in sparse decomposition to selectively determine said optimal combination of atoms for each said mutual comparison.

11. The system as recited in claim 10, wherein:

said SVM executes pair-wise comparisons of distinct source organisms, said SVM determining for each said pair-wise comparison of source organisms a two-dimensional decision subspace defined by a corresponding pair of optimal atoms; and

,

said classifier portion executes a non-parametric voting process iteratively mapping corresponding portions of said anatomic data segment sparse decomposition to each said decision subspace.

12. The system as recited in claim 11, wherein said simultaneous sparse approximation is carried out according to a greedy adaptive decomposition (GAD) process.

13. A system for taxonomically distinguishing anatomic samples captured in unconstrained manner for a plurality of source organisms, comprising:

a plurality of field acquisition portions for acquiring anatomic samples of the source organisms;

a portable imagery portion disposed at least at one of said field acquisition portion, said portable imagery portion actuating to capture images of the anatomic samples acquired at said field acquisition portion;

a mobile communication device coupled to said portable imagery portion, said mobile communication device actuating transmission of the captured images as anatomic image data segments to a classifier portion remotely disposed from said field acquisition portions;

said classifier portion actuating a server to execute;

a vector operation constructing for each of the anatomic image data segments at least one vector defined by pixel values for a predetermined strip of pixels across a corresponding one of the captured images, at least one of the anatomic image data segments being selected from a plurality of different versions of a captured image generated at different image resolutions and preserved for selection exclusive of one or more other generated versions;

a sparse decomposition operation selectively executing in at least a training system mode a simultaneous sparse approximation upon a joint corpus of pixel value vectors for a plurality of anatomic image data segments of distinct source organisms and matching image resolution, said sparse decomposition operation adaptively generating at least one sparse decomposition for each said vector with respect to a representative set of decomposition atoms;

a discriminant reduction operation executable during the training system mode to derive an optimal combination of atoms from said representative set for cooperatively distinguishing image signals attributable to the anatomic samples of different ones of the distinct source organisms; and

,

a classification operation executable in a classification system mode to discover for said sparse decomposition of an input anatomic image data segment a degree of correlation relative to the anatomic samples of each of the distinct source organisms; and

,

a database portion pre-stored with a plurality of anatomic data segments attributable respectively to a plurality of distinct source organisms, said classifier portion in the training system mode reading the pre-stored anatomic data segments from said database portion, said sparse decomposition and discriminant reduction operations of said classifier portion thereby executing in the training system mode to identify a distinct class corresponding to at least one of the known distinct source organisms;

wherein said classifier portion actuates a responsive transmission back to the mobile communication device originating an anatomic data segment transmission, the responsive transmission indicating a distinct source organism correlation discovered for the transmitted anatomic data segment,

different source organisms being thereby distinguished from one another based on anatomic data captured without constraint on the nature and form of data content.

14. The system as recited in claim 13, wherein at least one of said portions of the system is implemented as an online application for said mobile communication device for identifying and classifying the anatomic sample.

15. The system as recited in claim 14, wherein said classifier portion filters the distinct source organism correlation discovered for the anatomic data segment transmission according to at least one of:

pre-stored data on the distinct source organism and geographic location data for the field acquisition portions.

16. The system as recited in claim 14, wherein said portable imagery portion includes a manipulator base supporting at least one of an anatomic sample and an imaging device for angularly displacement of one relative to the other, said portable imagery portion thereby capturing the images of the anatomic sample in angularly adjustable manner.

17. The system as recited in claim 14, wherein:

said field acquisition portions include facilities for collecting the anatomic samples from at least one of;

vehicle grills, vehicle windshields, cargo articles, freight containers, naturally growing plant materials, stored specimen depositories, specimen collection traps, clothing of travelers, and luggage of travelers;

the anatomic samples include at least one of;

an insect wing image type and an arachnid profile image type; and

,

said vector operation of said classifier portion includes selective down-sampling of the anatomic sample images in resolution to construct the vectors of predetermined form for the anatomic data segments.

18. The system as recited in claim 17, wherein said discriminant reduction operation of said classifier portion includes a Support Vector Machine (SVM) programmably implemented therein, said SVM executing to mutually n-wise compare the distinct source organisms in sparse decomposition to selectively determine said optimal combination of atoms for each said mutual comparison.

19. The system as recited in claim 18, wherein:

said SVM executes pair-wise comparisons of distinct source organisms, said SVM determining for each said pair-wise comparison of source organisms a two-dimensional decision subspace defined by a corresponding pair of optimal atoms;

said classifier portion executes a non-parametric voting process iteratively mapping corresponding portions of said anatomic data segment sparse decomposition to each said decision subspace; and

,

said simultaneous sparse approximation is carried out according to a greedy adaptive decomposition (GAD) process.

20. A system for taxonomically distinguishing anatomic samples captured in unconstrained manner for a plurality of insect organisms, comprising:

a portable imagery portion actuating to capture images of anatomic samples acquired in at least one of a wing image and dorsal image form; and

,

a personal communication device coupled to said portable imagery portion, said personal communication device actuating transmission of the captured images as anatomic image data segments to a classifier portion remotely disposed from said portable imagery portion;

said classifier portion actuating a server to;

construct for each of the anatomic image data segments at least one vector defined by pixel values for a predetermined strip of pixels across a corresponding one of the captured images, at least one of the anatomic image data segments being selected from a plurality of different versions of a captured image generated at different image resolutions and preserved for selection exclusive of one or more other generated versions, and,

adaptively generate at least one decomposition for each vector of an anatomic image data segment with respect to a representative set of decomposition atoms, and discover for said decomposition of an input anatomic image data segment a degree of correlation relative to the anatomic samples of each of the distinct insect organisms,

different source organisms being thereby distinguished from one another based on anatomic data captured without constraint on the nature and form of data content.

21. A system for taxonomically distinguishing anatomic samples captured in unconstrained manner for a plurality of source organisms, comprising:

a portable imagery portion actuating to capture images of acquired anatomic samples of the source organisms;

a classifier portion remotely disposed from said portable imagery portion, and receiving the captured images as anatomic image data segments, said classifier portion actuating a server to execute;

a vector operation constructing for each of the anatomic image data segments at least one vector of predetermined form, at least one of the anatomic image data segments being selected from a plurality of different versions of a captured image generated at different image resolutions and preserved for selection exclusive of one or more other generated versions;

a sparse decomposition operation selectively executing in at least a training system mode a simultaneous sparse approximation upon a joint corpus of pixel value vectors for a plurality of anatomic image data segments of distinct source organisms and matching image resolution, said sparse decomposition operation adaptively generating at least one sparse decomposition for each said vector with respect to a representative set of decomposition atoms;

a discriminant reduction operation executable during the training system mode to derive an optimal combination of atoms from said representative set for cooperatively distinguishing image signals attributable to the anatomic samples of different ones of the distinct source organisms; and

,

a classification operation executable in a classification system mode to discover for said sparse decomposition of an input anatomic image data segment a degree of correlation in the optimal combination of atoms relative to the anatomic samples of each of the distinct source organisms,

different source organisms being thereby distinguished from one another based on anatomic data captured without constraint on the nature and form of data content.