Method for extracting features of a thermal image

Patent metadata

Field	Value
Publication number	US-10621466-B2
Application number	US-201816059051-A
Country	US
Kind code	B2
Filing date	Aug 9, 2018
Priority date	Nov 30, 2017
Publication date	Apr 14, 2020
Grant date	Apr 14, 2020

Abstract

A method for extracting features of a thermal image is provided. The method includes: reading a thermal image, and dividing the thermal image into a plurality of block images; and extracting a histogram of oriented gradient (HOG) feature histogram from each of the plurality of block images, and transforming the HOG feature histogram of each of the plurality of block images into a symmetric weighting HOG (SW-HOG) feature histogram. The SW-HOG feature histogram is obtained by multiplying a histogram of gradient intensity distribution by a block weighting. The method increases weightings of blocks which cover human contours and reduces weightings of blocks of an internal region of a human appearance through analyzing thermal lightness difference of regions within blocks, to reduce the influence of clothes in the internal region and the influence of the background region.

First claim

What is claimed is:

1. A method for extracting features of a thermal image, comprising: reading the thermal image, and dividing the thermal image into a plurality of block images; and extracting a histogram of oriented gradient (HOG) feature histogram from each of the plurality of block images, and transforming the HOG feature histogram of each of the plurality of block images into a symmetric weighting HOG (SW-HOG) feature histogram; wherein the SW-HOG feature histogram is obtained by multiplying a histogram of gradient intensity distribution by a block weighting, and the block weighting is:

w
⁡

(

B
i

)

=

d
⁡

(

B
i

)

∑

j
=
1

9

⁢

(

c
i

l
,
t

⁡

(
j
)

+

c
i

l
,
b

⁡

(
j
)

+

c
i

r
,
t

⁡

(
j
)

+

c
i

r
,
b

⁡

(
j
)

)

wherein B i represents a block image within the plurality of block images, w(B i ) represents the block weighting of the block image B i , d(B i ) represents gradient intensity of the block image B i , and {c i l,t (j), c i l,b (j), c i r,t (j), c i r,b (j)} represent the j th bin of intensity of a top-left corner cell image, a bottom-left corner cell image, a top-right corner cell image and a bottom-right corner cell image, respectively.

2. The method of claim 1 , wherein each of the plurality of block images is divided into four cell images.

3. The method of claim 1 , wherein the HOG feature histogram is the histogram of gradient intensity distribution.

4. The method of claim 3 , wherein the histogram of gradient intensity distribution is obtained by calculating a histogram of horizontal gradient intensity distribution and a histogram of vertical gradient intensity distribution.

5. The method of claim 1 , wherein a magnitude of the block weighting is adjusted according to symmetry of the HOG feature histogram.

Assignees

• Nat Chung Shan Inst Science & Tech

Inventors

• Huang Shih-Shinh
• Chien Shih-Che
• Chang Feng-Chia
• Hsiao Chien-Hao
• Hsiao Yu-Sung

Classifications

• G06K9/00362

  Physics · mapped topic

• G06K9/4647 Primary

  Physics · mapped topic

• G06K9/00369

  Physics · mapped topic

• G06V40/103 Primary

  Static body considered as a whole, e.g. static pedestrian or occupant recognition · CPC title

• G06V10/507

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