Automatic visual recognition of biological particles

• US 20050251347A1
• Filed: 05/05/2005
• Published: 11/10/2005
• Est. Priority Date: 05/05/2004
• Status: Abandoned Application

• Current Assignee: Marc'Aurelio Ranzato, Pietro Perona, Richard C. Flagan
• Original Assignee: Marc'Aurelio Ranzato, Pietro Perona, Richard C. Flagan
• Inventors: Ranzato, Marc'aurelio, Perona, Pietro, Flagan, Richard C.
• Application Number:
  US11/122,575
• Publication Number:
  US 20050251347A1
• Time in Patent Office: Days
• US Class Current: 702/19
• CPC Class Codes: G06F 18/24155 Bayesian classification
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: G06V 20/693 Acquisition

Abstract

A method and system provide the ability to automatically recognize biological particles. An image of biological particles (e.g., airborne pollen or urine) is obtained. One or more parts of the image are detected as containing one or more particles of interest. Feature vector(s) are extracted from each detected part of the image. Non-linearities are applied to each feature vector. Each part of the image is then classified into a category of biological particle based on the one or more feature vectors for each part of the image.

Claims

1. A method for recognizing biological particles, comprising:

obtaining an image comprising biological particles;

detecting one or more parts of the image as containing one or more particles of interest;

extracting one or more feature vectors from each detected part of the image;

applying one or more non-linearities to each feature vector; and

classifying each part of the image into a category of biological particle based on the one or more feature vectors for each part of the image.

2. The method of claim 1, wherein the image is obtained using a volumetric spore trap and comprises images of airborne pollen.

3. The method of claim 1, wherein the image is obtained using a light microscope and comprises images of urine.

4. The method of claim 1, wherein the detecting is based on a filtering approach using a difference of Gaussians (DoG).

5. The method of claim 1, wherein the detecting provides a part of the image that is invariant with respect to scale, shift, and rotation.

6. The method of claim 1, wherein one of the non-linearities is applied to an invariant and comprises:

a piece-wise linear function; and

a piece-wise quadratic transformation that depends on a range of input invariants.

7. The method of claim 1, wherein one of the non-linearities is applied to an invariant and comprises dividing a range of variation of a signal of the invariant in each dimension into three parts wherein one part is sensitive to low values, a second part is sensitive to values around a background mean, and a third part is sensitive to high values.

8. The method of claim 1, wherein one of the non-linearities is applied to one or more invariants and comprises:

dividing each invariant into three parts—

positive, negative, and absolute value;

adding a background mean to the absolute value.

9. A system for recognizing biological particles comprising:

(a) an image of biological particles;

(b) a detector configured to detect one or more parts of the image as containing one or more particles of interest; and

(c) a classifier configured to;

(i) extract one or more feature vectors from each detected part of the image;

(ii) apply one or more non-linearities to each feature vector; and

(iii) classify each part of the image into a category of biological particle based on the one or mote feature vectors for each part of the image.

10. The system of claim 9, further comprising a volumetric sport trap configured to obtain the image wherein the image comprises images of airborne pollen.

11. The system of claim 9, further comprising a light microscope configured to obtain the image wherein the image comprises images of urine.

12. The system of claim 9, wherein the detector is configured to detect one or more parts of the image based on a filtering approach using a difference of Gaussians (DoG).

13. The system of claim 9, wherein the detector is configured to provide a part of the image that is invariant with respect to scale, shift, and rotation.

14. The system of claim 9, wherein one of the non-linearities is applied to an invariant and comprises:

a piece-wise linear function; and

a piece-wise quadratic transformation that depends on a range of input invariants.

15. The system of claim 9, wherein one of the non-linearities is applied to an invariant and comprises dividing a range of variation of a signal of the invariant in each dimension into three parts wherein one part is sensitive to low values, a second part is sensitive to values around a background mean, and a third part is sensitive to high values.

16. The system of claim 9, wherein one of the non-linearities is applied to one or more invariants and comprises:

dividing each invariant into three parts—

positive, negative, and absolute value;

adding a background mean to the absolute value.