Method and Apparatus for Detecting Targets

Patent metadata

Field	Value
Publication number	US-2016371850-A1
Application number	US-201514743297-A
Country	US
Kind code	A1
Filing date	Jun 18, 2015
Priority date	Jun 18, 2015
Publication date	Dec 22, 2016
Grant date	—

Abstract

A method and apparatus for performing target detection. A potential object is detected within a candidate chip in an image. The potential object is verified as a candidate object. The candidate object is classified as one of a candidate target or a background in response to the potential object being verified as the candidate object. The candidate target is verified as a target of interest in response to the candidate object being classified as the candidate target.

First claim

What is claimed is:

1 . A method for performing target detection, the method comprising:
detecting a potential object within a candidate chip in an image; verifying that the potential object is a candidate object; classifying the candidate object as one of a candidate target or a background in response to the potential object being verified as the candidate object; and verifying that the candidate target is a target of interest in response to the candidate object being classified as the candidate target.

2 . The method of claim 1 further comprising:
generating a target signature for the target of interest; and
tracking the target of interest in a set of subsequent images using the target signature.

3 . The method of claim 1 , wherein verifying that the candidate target is the target of interest comprises:
identifying a plurality of sub-regions for the candidate chip using a number of sub-region masks; generating a plurality of scores for the plurality of sub-regions; and assigning each of the plurality of sub-regions to one of a target class and a background class based on the plurality of scores for the plurality of sub-regions.

4 . The method of claim 3 , wherein verifying that the candidate target is the target of interest further comprises:
determining whether a ratio of a first parameter computed for the target class to a second parameter computed for the background class is at least a selected weighted ratio; and verifying, successfully, that the candidate target is the target of interest with a desired level of accuracy in response to a determination that the ratio is at least the selected weighted ratio.

5 . The method of claim 4 , wherein verifying that the candidate target is the target of interest further comprises:
computing a new ratio in response to a determination that the ratio is less than the selected weighted ratio; determining whether the new ratio is less than a differentiation factor; verifying, successfully, that the candidate target is the target of interest with a desired level of accuracy in response to a determination that the new ratio is less than the differentiation factor; and identifying the candidate target as the background in response to a determination that the new ratio is not less than the differentiation factor.

6 . The method of claim 1 , wherein verifying that the potential object is the candidate object comprises:
determining whether the potential object matches a previously detected potential object in at least one previous image; and updating a consistency index associated with a unique identifier assigned to the previously detected potential object in response to a determination that the potential object matches the previously detected potential object in the at least one previous image.

7 . The method of claim 6 , wherein verifying that the potential object is the candidate object further comprises:
determining whether the consistency index has reached a selected threshold value; and identifying the potential object as the candidate object in response to a determination that the consistency index has reached the selected threshold value.

8 . The method of claim 6 , wherein verifying that the potential object is the candidate object further comprises:
assigning a new unique identifier to the potential object in response to a determination that the potential object does not match any previously detected potential object.

9 . The method of claim 6 , wherein determining whether the potential object matches the previously detected potential object in the at least one previous image comprises:
identifying an appearance of the potential object in the image using at least one of a size, a chrominance pattern, a color pattern, or a shape of the potential object; and determining whether the potential object matches the previously detected potential object in the at least one previous image based on the appearance of the potential object.

10 . The method of claim 6 , wherein determining whether the potential object matches the previously detected potential object in the at least one previous image comprises:
identifying a direction of movement for the potential object in the previous image; and determining whether the potential object matches the previously detected potential object in the at least one previous image based on the direction of movement for the potential object.

11 . The method of claim 1 , wherein classifying the candidate object comprises:
classifying the candidate object as one of the candidate target or the background using rotationally shifted histograms of oriented gradients in response to the potential object being verified as the candidate object.

12 . The method of claim 1 , wherein detecting the potential object comprises:
detecting a potential moving object within the candidate chip in the image using a moving object detector.

13 . A method for processing images, the method comprising:
detecting a potential moving object within a candidate chip in an image; verifying that the potential moving object is a candidate object based on at least one of an appearance of the potential moving object or a direction of movement of the potential moving object; classifying the candidate object as one of a candidate target or a background in response to the potential moving object being verified as the candidate object; and verifying that the candidate target is a target of interest in response to the candidate object being classified as the candidate target using a number of reference target chips.

14 . The method of claim 13 , wherein verifying that the candidate target is the target of interest comprises:
identifying a plurality of sub-regions for the candidate chip using a number of sub-region masks; generating a plurality of scores for the plurality of sub-regions; assigning each of the plurality of sub-regions to one of a target class and a background class based on the plurality of scores for the plurality of sub-regions; and verifying that the candidate target is the target of interest using the number of reference target chips, the target class, and the background class.

15 . An apparatus comprising:
an object detector that detects a potential object within a candidate chip in an image and verifies that the potential object is a candidate object; and a classifier that classifies the candidate object as one of a candidate target or a background in response to the potential object being verified as the candidate object and verifies that the candidate target is a target of interest in response to the candidate object being classified as the candidate target.

16 . The apparatus of claim 15 , wherein the object detector comprises:
a first verifier that verifies that the potential object is the candidate object based on at least one of an appearance of the potential object or a direction of movement of the potential object.

17 . The apparatus of claim 16 , wherein the appearance is identified using at least one of a size, a chrominance pattern, a color pattern, or a shape of the potential object.

18 . The apparatus of claim 15 , wherein the classifier comprises:
a second verifier that identifies a plurality of sub-regions for the candidate chip using a number of sub-region masks, generates a plurality of scores for the plurality of sub-regions, assigns each of the plurality of sub-regions to one of a target class and a background class based on the plurality of scores for the plurality of sub-regions, and verifie

Assignees

• Boeing Co

Inventors

• Kwon Hyukseong
• Kim Kyungnam
• Owechko Yuri

Classifications

• G06V10/255 Primary

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

• G06F18/24

  Classification techniques · CPC title

• G06F18/22

  Matching criteria, e.g. proximity measures · CPC title

• G06V10/768

  using context analysis, e.g. recognition aided by known co-occurring patterns · CPC title

• G06V10/507

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