Segmentation method for use against moving objects

• US 5,109,435 A
• Filed: 03/04/1991
• Issued: 04/28/1992
• Est. Priority Date: 08/08/1988
• Status: Expired due to Fees

• Current Assignee: Hughes Aircraft Company (Rtx Corporation)
• Original Assignee: Hughes Aircraft Company (Rtx Corporation)
• Inventors: Sacks, Jack M., Lo, Thomas K., Banh, Nam D.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): CAMMARATA, MICHAEL
• Application Number:
  US07/663,786
• Time in Patent Office: 421 Days
• Field of Search: 382/48, 382/50, 382/52, 382/1, 358/105, 358/108, 358/125, 358/126, 358/139, 342/93, 342/351
• US Class Current: 382/103
• CPC Class Codes: G06F 17/153 Multidimensional correlation or convolution
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: H03H 17/0263 Rank order filters

Abstract

Three image frames containing an object of interest and background clutter are taken at successive time intervals and stored in memory. The background of images A, B and C are registered preferably using an area correlator 12. A median filter 16 is used to select a median value from the registered image frames. Then, subtractor 18 serves to subtract the median pixel values from one of the image frames. This difference output is then thresholded to provide a binary signal whose pixel values exceeding the threshold levels are generally associated with the position of the moving object.

Claims

1. A method of detecting a moving object in a plurality of image frames taken at different times, each image frame containing the object and background clutter, the method comprising the steps of:

a) registering the background of at least three image frames;

b) determining a median pixel value for each pixel position of the registered image frames to construct a registered image frame representative of only the background of the three image frames without the object;

c) subtracting the median pixel values from pixels in a selected one of the image frames to form a difference image; and

d) applying a threshold to the difference image whereby pixel values exceeding a given threshold level are associated with the position of the moving object.

2. The method of claim 1 wherein the threshold level is fixed at a given multiple of noise in the image frames.

3. The method claim 1 wherein the difference image is thresholded against positive and negative threshold levels, with pixel values above the positive threshold level generally being associated with bright portions of the object while pixel values below the negative threshold are generally associated with dark portions of the object.

4. The method of claim 3 which further comprises:

calculating an aimpoint from the combined bright and dark portions of the detected object.

5. The method of claim 4 wherein the aimpoint is calculated from the centroid of the combined bright and dark portions of the object.

6. The method of claim 1 wherein said selected image frame is an image frame most recently taken from a sensor.

7. The method of claim 1 wherein step a) is performed using an area correlator.

8. The method of claim 1 wherein the difference image is applied to two pairs of positive and negative threshold levels, one pair of thresholds being scaled for optimum object of interest detection while another pair is scaled so as to lie as close to the noise level in the image frames.

9. A method of detecting moving objects in cluttered images, said method comprising the steps of:

a) sensing a first scene at a given time which contains the moving object and background and storing the first scene in a memory;

b) sensing substantially the same scene at a subsequent time and storing it as a second image frame in a memory;

c) sensing substantially the same scene and storing it as a third image frame in a memory;

d) using an area correlator to substantially register the image frames;

e) selecting a median pixel value for each pixel position of the registered image frames to construct a registered image frame representative of only the background of the of three image frames without the object;

f) subtracting the median pixel values from pixel values in the third image frame to thereby form a difference image; and

g) generating a binary output signal by thresholding the difference image against preselected positive and negative threshold levels chosen as a function of noise in the image frames whereby pixel values in the difference image exceeding the positive and negative threshold values are generally associated with dark and bright portions of the moving object, respectively.

10. The method of claim 9 which further comprises:

calculating an aimpoint from the centroid of the dark and bright portions of the moving object as represented in the binary output signal.

11. Apparatus for detecting a moving object in a plurality of image frames taken at different times, each image frame containing the object and background clutter, the apparatus comprising:

means for registering the background of at least three image frames;

means for determining a median pixel value for each pixel position of the registered images to construct a registered image frame representative of only the background of the three image frames without the object;

means for subtracting the median pixel values from pixels in a selected one of the image frames to form a difference image; and

means for applying a threshold to the difference image whereby pixel values exceeding a given threshold level are associated with the position the moving object.

12. The apparatus of claim 11 which further comprises:

means for defining a positive threshold level;

means for defining a negative threshold level; and

wherein the difference image is thresholded against the positive and negative threshold levels, with pixel values above the positive threshold levels generally being associated with bright portions of the object while pixel values below the negative threshold are generally associated with dark portions of the object.

13. The apparatus of claim 12 which further comprises:

means for calculating an aimpoint from the combined bright and dark portions of the detected object.

14. The apparatus of claim 13 wherein the means for calculating includes:

means for calculating the centroid of the bright portion of the object; and

means for calculating the centroid of the dark portion of the object.

15. The apparatus of claim 11 wherein said means for registering comprises an area correlator.

16. The apparatus of claim 11 wherein said means for applying a threshold comprises:

first positive threshold means for defining a first positive threshold level that is scaled to optimize detection of the object of interest;

first negative threshold means for defining a negative threshold level that is scaled to optimize detection of the object of interest;

second positive threshold means for defining a second positive threshold level close to the noise level in the image frames; and

second negative threshold means for defining a second negative threshold level that is scaled close to the noise level in the image frames.

17. The apparatus of claim 16 wherein said second positive threshold means and second negative threshold means define threshold levels which are a given multiple of the noise in the image frames.