Tracking moving objects accurately on a wide-angle video

• US 8,139,896 B1
• Filed: 03/28/2006
• Issued: 03/20/2012
• Est. Priority Date: 03/28/2005
• Status: Expired due to Fees

• Current Assignee: Grandeye Limited (Oncam Global Group AG)
• Original Assignee: Grandeye Limited (Oncam Global Group AG)
• Inventors: Ahiska, Yavuz, Castellari, Andrea Elvis, Davey, Mark Kenneth
• Primary Examiner(s): Bhatnagar, Anand
• Assistant Examiner(s): BLOOM, NATHAN J
• Application Number:
  US11/390,957
• Time in Patent Office: 2,184 Days
• Field of Search: 382/103, 382/190, 382/195, 382203-204, 382224-225, 382/275, 382/286, 382/217, 348/143, 348/147, 348152-154, 348/159
• US Class Current: 382/298
• CPC Class Codes: G06T 3/12 Panospheric to cylindrical image transformations
• CPC Class Codes: G06V 20/53 Recognition of crowd images, e.g. recognition of crowd congestion

Abstract

A method and a system for tracking the motion of moving objects accurately on the entirety of a wide-angle video is disclosed. The method includes using a non-uniform scaling to selectively enhance pixel density, preferably in preparation for other image processing. In preferred embodiments, the further image processing (such as motion detection, object recognition, or tracking, etc.) functions better with the enhanced pixel density or distribution.

Claims

1. A method for performing image analysis functions in image processing circuitry, comprising:

(a) capturing a sequence of high-resolution frames through a wide-angle non-perspective optical train;

(b) reducing the pixel resolution of at least some frames in said sequence to thereby produce reduced resolution frames, using a procedure which reduces a pixel count of selected transitory image portions corresponding to imaged objects, wherein the selected image portions corresponding to small-appearing objects are reduced in resolution less than the selected image portions corresponding to large-appearing objects, wherein the reduction of resolution results in a closer similarity in the size of the selected image portions corresponding to transitory objects and maintains at least a minimum number of pixels for the selected image portions; and

(c) performing an image analysis that requires the minimum number of pixels on said reduced resolution frames, and outputting the result of said image analysis.

2. The method of claim 1, wherein the reduced resolution image undergoes at least one of:

motion detection analysis;

object detection analysis;

object recognition analysis;

object tracking analysis; and

object behavior analysis.

3. The method of claim 1, wherein said image analysis occurs on an external device.

4. The method of claim 1, wherein the reduced resolution frames exhibit a different aspect ratio in at least some parts of the frames compared to the original high-resolution frames.

5. A method of detecting and tracking moving objects in a captured wide-angle digital video sequence, comprising:

capturing a sequence of wide-angle high-resolution images using a wide-angle digital video camera;

selecting image parts in at least some of said high-resolution images, said selected parts corresponding to one or more transitory objects of interest in said high-resolution images;

converting at least some of said high-resolution images into reduced-size lower-resolution images, selectively scaling down resolution of said selected image parts differently, based on the relative apparent size of the transitory objects of interest, using a spatially non-linear transform, to maintain a minimum pixel count in said objects of interest; and

performing an image analysis function on said lower-resolution images, wherein said pixel count is defined by said image analysis function.

6. The method of claim 5, wherein said pixel count comprises a threshold for operation of said image analysis function.

7. The method of claim 5, wherein the reduced-size lower resolution image is digitally transmitted to an external device which receives and performs image analysis on said reduced-size lower resolution image.

8. The method of claim 5, wherein maintaining said pixel count produces visually distorted objects with less proportionally dissimilar pixel counts than corresponding objects in said high-resolution images, altering the aspect ratio in parts of the lower-resolution images containing the objects compared to corresponding parts in the high-resolution images.

9. The method of claim 5, wherein the lower-resolution images undergo at least one of:

motion detection analysis;

object detection analysis;

object recognition analysis;

object tracking analysis; and

object behavior analysis.

10. A method for performing image analysis of a video image sequence, comprising:

capturing a sequence of wide-angle non-perspective frames with optically distorted images;

performing nonlinear reduction of a pixel count of smaller transitory image portions disproportionately to a pixel count of larger transitory image portions, within the same distorted image, in at least some captured frames to produce reduced resolution frames, while selectively preserving at least a first pixel count for the transitory image portions, the first pixel count being defined for optimization of an image analysis function;

performing the image analysis function on said reduced resolution frames; and

outputting a result of said image analysis.

11. The method of claim 10, wherein the reduced resolution frames undergo at least one of:

motion detection analysis;

object detection analysis;

object recognition analysis;

object tracking analysis; and

object behavior analysis.

12. The method of claim 10, wherein the first pixel count comprises a threshold for performance of said image analysis.

13. The method of claim 11, wherein the selected portions of the reduced resolution frames exhibit a different aspect ratio compared to corresponding portions of the captured frames.