Method and system for diet management

Patent metadata

Field	Value
Publication number	US-9449029-B2
Application number	US-201313854970-A
Country	US
Kind code	B2
Filing date	Apr 2, 2013
Priority date	Dec 14, 2012
Publication date	Sep 20, 2016
Grant date	Sep 20, 2016

Abstract

A system and a method for diet management based on image analysis are provided. The system includes a database and a comparison device. The comparison device is coupled to the database. The comparison device performs similarity comparison in the database based on a supervector related to at least one diet image so as to find out at least one similar population and provides information related to the at least one similar population.

First claim

What is claimed is:

1. A computer-implemented method for diet management, adapted to a diet management system, comprising:
providing a supervector, wherein the supervector is related to at least one diet image;
performing similarity comparison in a database based on the supervector by a comparison device, so as to find out at least one similar population; and
providing information related to the similar population,
wherein the providing the information related to the similar population comprises:
calculating at least one statistic of the similar population by using at least one multidimensional data of the similar population; and
providing the information related to the similar population based on the statistic of the similar population.

2. The method for diet management of claim 1 , wherein the providing the supervector comprises:
capturing the at least one diet image via an image capture device;
pre-processing the at least one diet image so as to obtain at least one detailed food segment;
extracting at least one diet image feature from the detailed food segment; and
generating the supervector based on the at least one diet image feature.

3. The method for diet management of claim 2 , wherein the pre-processing the at least one diet image comprises:
transforming the at least one diet image to a normalized space.

4. The method for diet management of claim 2 , wherein the pre-processing the at least one diet image comprises:
performing color correction, bright correction, or white balance correction on the at least one diet image; and
obtaining the at least one detailed food segment from the at least one diet image.

5. The method for diet management of claim 4 , wherein the obtaining the at least one detailed food segment from the at least one diet image comprises:
performing background removal on the at least one diet image so as to obtain at least one diet region; and
performing detailed food segment segmentation on the at least one diet region so as to segment out the at least one detailed food segment from the diet region.

6. The method for diet management of claim 2 , wherein the at least one diet image feature comprises image capturing time, image capturing location, color, texture complexity, or reflectivity.

7. The method for diet management of claim 6 , wherein the texture complexity comprises texture magnitude, texture orientation, texture length, texture amount, or texture regularity.

8. The method for diet management of claim 2 , wherein the extracting the at least one diet image feature comprises:
performing an X-direction difference calculation and a Y-direction difference calculation on the detailed food segment so as to obtain an X-direction difference image and a Y-direction difference image;
calculating texture magnitude information and texture orientation information by using the X-direction difference image and the Y-direction difference image;
calculating at least one texture coding in at least one direction in the detailed food segment by using the texture magnitude information and the texture orientation information; and
integrating the at least one texture coding in the at least one direction so as to obtain the at least one diet image feature.

9. The method for diet management of claim 8 , wherein the calculating the texture magnitude information and the texture orientation information comprises:
calculating the texture magnitude information e(x,y)=√{square root over (g x 2 +g y 2 )}, wherein gx represents a pixel value of the X-direction difference image at (x,y), and gy represents a pixel value of the Y-direction difference image at (x,y); and
calculating the texture orientation information

θ
⁡

(

x
,
y

)

=

tan

-
1

⁡

[

g
y

g
x

]

.

10. The method for diet management of claim 8 , wherein the calculating the at least one texture coding in the at least one direction in the detailed food segment comprises:
selecting a target pixel from the detailed food segment;
selecting a coding region along the at least one direction in the detailed food segment, wherein the coding region comprises the target pixel and a plurality of neighboring pixels;
performing binarization on the texture orientation information of the neighboring pixels in the coding region so as to obtain binary values of the neighboring pixels;
transforming the binary values of the neighboring pixels to a category value of the target pixel;
determining a height value of the target pixel based on the texture magnitude information of the target pixel; and
determining the at least one texture coding in the at least one direction in the detailed food segment based on the category values and the height values in the detailed food segment.

11. The method for diet management of claim 10 , wherein the performing binarization on the texture orientation information of the neighboring pixels in the coding region comprises:
selecting a neighboring pixel from the neighboring pixels;
the binary value of the neighboring pixel being 1 if |θi−θn|≦γθ, wherein θi represents the texture orientation information of the neighboring pixel, and wherein θn represents an angle of a normal vector of the at least one direction, and wherein γθ represents an angle threshold value; and
the binary value of the neighboring pixel being 0 if |θi−θn|<γθ.

12. The method for diet management of claim 10 , wherein the transforming the binary values of the neighboring pixels to the category value of the target pixel comprises:
calculating the category value

bin
=

1
+

∑

t
=
1

m

⁢

⁢

2

m
-
t

×

D
⁡

(
t
)

,

wherein m represent the number of the neighboring p

Assignees

• Ind Tech Res Inst

Inventors

• Ting Chuan-Wei
• Wang Ching-Yao
• Chen Ju-Chin

Classifications

• G06V20/68

  Food, e.g. fruit or vegetables · CPC title

• G06V10/507

  Summing image-intensity values; Histogram projection analysis · CPC title

• G16H30/20

  for handling medical images, e.g. DICOM, HL7 or PACS · CPC title

• G16H30/40

  for processing medical images, e.g. editing · CPC title

• G09B19/0092

  Nutrition · CPC title