Correlation filters for target reacquisition in trackers

• US 5,947,413 A
• Filed: 09/05/1997
• Issued: 09/07/1999
• Est. Priority Date: 11/12/1996
• Status: Expired due to Term

• Current Assignee: Hughes Electronics Corporation (Rtx Corporation), Raytheon Company (Rtx Corporation)
• Original Assignee: Raytheon Company (Rtx Corporation)
• Inventors: Mahalanobis, Abhijit
• Primary Examiner(s): Jordan, Charles T.
• Assistant Examiner(s): Wesson, Theresa M.
• Application Number:
  US08/924,396
• Time in Patent Office: 732 Days
• Field of Search: 244/3.16, 244/3.17, 382/278, 382/103, 701/223
• US Class Current: 244/3.17
• CPC Class Codes: G01S 3/781 Details
• CPC Class Codes: G01S 3/7865 using correlation of the live video image with a stored image

Abstract

A system and method for target reacquisition and aimpoint selection in missile trackers. At the start of iterations through the process, distance classifier correlation filters (DCCFs) memorize the target'"'"'s signature on the first frame. This stored target signature is used in a subsequent confidence match test, so the current sub-frame target registration will be compared against the stored target registration from the first frame. If the result of the match test is true, a patch of image centered on the aimpoint is used to synthesize the sub-frame filter. A sub-frame patch (containing the target) of the present frame is selected to find the target in the next frame. A next frame search provides the location and characteristics of a peak in the next image, which indicates the target position. The DCCP shape matching processing registers the sub-frame to the lock coordinates in the next frame. This process will track most frames, and operation will repeat. However, when the similarity measure criterion is not satisfied, maximum average correlation height (MACH) filters update the aim-point and re-designate the track-point. Once the MACH filters are invoked, the process re-initializes with the new lock coordinates. The MACH filters have pre-stored images which are independent of target and scene data being processed by the system.

Claims

1. A method for tracking the position of a target in a sequence of image frames provided by a sensor, comprising a sequence of the following steps:

(i) receiving an initial image frame with a preliminary set of target position data indicating a preliminary target position in relation to the image frame;

(ii) processing the image frame with a distance classifier correlation filter (DCCF) to register a sub-frame patch of said image frame using said set of target position data, wherein the target is centered in relation to said sub-frame patch;

(iii) using said sub-frame patch to update a sub-frame filter used for searching a successive image frame for an updated location of said target in said successive image frame and providing an updated set of target position data;

(iv) receiving a subsequent image frame and processing said subsequent image frame with said DCCF to register a successive sub-frame patch of said successive image frame, wherein the target is centered in relation to said successive sub-frame patch using said updated set of target position data; and

(v) under a preselected set of conditions, performing an independent determination of the target position using a maximum average correlation height (MACH) filter to provide a fresh set of target position data.

2. The method of claim 1 wherein said preselected set of conditions includes the condition that the sub-frame patch registered by the DCCF for initial first image frame does not match with a sub-frame patch registered by the DCCF for a subsequent image frame using a similarity measure.

3. The method of claim 1 further comprising the step of testing said successive sub-frame patch of said successive image frame to determine a measure of the goodness of match between the sub-frame patch of said initial image frame and said successive sub-frame patch, repeating steps (iii) and (iv) if a good match according to said measure is found, and performing step (v) if a good match according to said measure is not found.

4. A method for tracking the position of a target in a sequence of image frames provided by a sensor, comprising a sequence of the following steps:

(i) receiving an initial image frame with a preliminary set of target position data indicating a preliminary target position in relation to the image frame;

(ii) processing the image frame with a distance classifier correlation filter (DCCF) to register a sub-frame patch of said image frame using said set of target position data, wherein the target is centered in relation to said sub-frame patch;

(iii) using said sub-frame patch to update a sub-frame filter used for searching a successive image frame for an updated location of said target in said successive image frame and providing an updated set of target position data;

(iv) processing the subsequent image frame with said DCCF to register a successive sub-frame patch of said successive image frame, wherein the target is centered in relation to said successive sub-frame patch using said updated set of target position data;

(v) testing successive sub-frame patch of said successive image frame to determine a measure of the goodness of match between the sub-frame patch of said initial image frame and said successive sub-frame patch, and repeating steps (iii) and (iv) if a good match according to said measure is found;

(vi) if a good match according to said measure is not found, performing an independent determination of the target position using a maximum average correlation height (MACH) filter to provide a fresh set of target position data; and

(vii) using the fresh set of target position data, repeating steps (iii) and (iv).

5. An airborne missile, comprising:

a sensor assembly for providing a sequence of image frames;

a navigation guidance and control apparatus; and

a tracker apparatus responsive to the sequence of image frames, said apparatus including apparatus responsive to the initial frame for providing a preliminary set of target position data indicating a preliminary target position in relation to the image frame, a distance classifier correlation filter (DCCF) to register a sub-frame patch of said image frame using said set of target position data, wherein the target is centered in relation to said sub-frame patch, apparatus using said sub-frame patch to update a sub-frame filter for searching a successive image frame for an updated location of said target in said successive image frame and providing an updated set of target position data, and apparatus for performing an independent determination of the target position under predetermined conditions, said apparatus comprising a maximum average correlation height (MACH) filter to provide a fresh set of target position data.