Multi-cue object detection and analysis

Patent metadata

Field	Value
Publication number	US-2016284097-A1
Application number	US-201615176289-A
Country	US
Kind code	A1
Filing date	Jun 8, 2016
Priority date	Jun 14, 2012
Publication date	Sep 29, 2016
Grant date	—

Abstract

Foreground objects of interest are distinguished from a background model by dividing a region of interest of a video data image into a grid array of individual cells. Each of the cells are labeled as foreground if accumulated edge energy within the cell meets an edge energy threshold, or if color intensities for different colors within each cell differ by a color intensity differential threshold, or as a function of combinations of said determinations.

First claim

What is claimed is:

1 . A computer-implemented method for distinguishing foreground objects of interest from a background model, the method comprising executing on a processing unit the steps of:
acquiring frame image data for each of a plurality of individual cells in a grid array that defines a region of interest of a video data image; for each of the cells, determining a first foreground value as true in response to determining that a color intensity of the acquired frame image data for the cell of one color is greater than a color intensity of the acquired frame image data of another color by a color intensity differential threshold; for each of the cells, determining a second foreground value as true in response to an accumulated energy of edges detected within the acquired frame image data for the cell meeting an edge energy threshold; labelling each of the cells as foreground in response to a true value output of a combination rule applied to the first foreground values and the second foreground values of the cells, wherein the combination rule is selected from the group consisting of a logical AND operation, a logical OR operation and a differentially weighted comparison operation; and using the frame image data from the cells labeled as foreground to define a foreground object.

2 . The method of claim 1 , further comprising:
integrating computer-readable program code into a computer system comprising the processing unit, a computer readable memory and a computer readable hardware storage device; and wherein the computer readable program code is embodied on the computer readable hardware storage device and comprises instructions for execution by the processing unit via the computer readable memory that cause the processing unit to perform the steps of acquiring the frame image data for each of the plurality of individual cells, determining the first foreground values and the second foreground values for each of the cells, labelling each of the cells as foreground in response to the true value output of the combination rule, and using the frame image data from the cells labeled as foreground to define the foreground object.

3 . The method of claim 1 , wherein the step of determining for each of the cells the second foreground value comprises:
generating a combination edge energy logical output value as an output of another combination rule applied to a plurality of accumulated energy outputs of multiple different edge detection processes for the each cell, wherein the another combination rule is selected from the group consisting of a logical AND operation and a logical OR operation; and determining the second foreground value as the generated combination edge energy logical output value.

4 . The method of claim 1 , wherein the combination rule of the step of labelling each of the cells as foreground is the differentially weighted comparison operation;
wherein the first and the second foreground values are differentially weighted; and wherein the step of labelling each of the cells as foreground is in response to a true value of a more heavily weighted one of the first and the second foreground values.

5 . The method of claim 1 , further comprising:
applying a Sobel operator to each cell to create an edge patch; and determining for each cell the second foreground value as true in response to determining a value of “1”, and as false in response to determining a value of “0,” according to:

f
e



(

c
m

)

=

{

1

∑

i
=
1

N



E


(

x
i

)

>

t
e

0

∑

i
=
1

N



E


(

x
i

)

≤

t
e

;

wherein c m is an m th cell of the cells, E(x i ) is an energy of the edge patch at pixel x i of an “N” plurality of pixels “i” of the cell c m , and cell c m is relabeled as “1” to signify foreground in response to determining that the summation of edge energy of the N pixels of the cell c m is bigger than an edge energy threshold t e .

6 . The method of claim 5 , wherein the different colors of the determined color intensities are one of red, relative green and blue color channels.

7 . The method of claim 6 , further comprising:
determining for each cell the first foreground value as true in response to determining a value of “1”, and as false in response to determining a value of “0”, according to:

f
c



(

c
m

)

=

{

0

min


(

Assignees

• IBM

Inventors

• Datta Ankur
• Feris Rogerio S
• Pankanti Sharathchandra U
• Wang Xiaoyu

Classifications

• G06F2218/12

  Classification; Matching · CPC title

• G06V10/50

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title

• G06V10/28

  Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns · CPC title

• G06V10/255

  Detecting or recognising potential candidate objects based on visual cues, e.g. shapes · CPC title

• G06K9/3241

  Physics · mapped topic