Method for detecting and tracking moving objects in a digital image sequence having a stationary background

• US 5,034,986 A
• Filed: 02/12/1990
• Issued: 07/23/1991
• Est. Priority Date: 03/01/1989
• Status: Expired due to Term

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Karmann, Klaus-Peter, von Brandt, Achim
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Johns, Andrew W.
• Application Number:
  US07/479,395
• Time in Patent Office: 526 Days
• Field of Search: 382/41, 382/48, 382/1, 358/126, 342/64, 342/189
• US Class Current: 382/103
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06V 10/24 Aligning, centring, orientation detection or correction of the image
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns

Abstract

A method for detecting and tracking moving objects in a digital image sequence having a stationary background by comparing the image sequence to a calculated background image sequence, whereby a sequence of binary object masks is calculated whose segments reproduce the shapes and positions of the moving subjects, and whereby the motion vectors of the moving objects are calculated by matching these segments. Also, a sequence of background images is calculated by spatially selective and time recursive averaging of the input image sequence in which the moving objects are not contained but in which other modifications of the background that are not caused by moving objects are contained. Further, a sequence of binary object masks is calculated by binarization of the difference image sequence from the input image sequence and the background image sequence using a threshold whose value is used for controlling the spatial selectivity in the calculation of the background image sequence and whose segments are determined together with their sizes and center of gravity positions. Finally, motion vectors of the detected objects are determined by minimization of the squared gray scale value differences averaged over the intersection of the shifted mask segments, whereby differences between the center of gravity vectors of corresponding segments are utilized as start vectors in the minimization.

Claims

1. Method for detecting and tracking moving objects in a digital input image sequence having a substantially stationary background by comparing the input image sequence to at least one calculated background image sequence, comprising:

calculating a sequence of binary object masks by binarization of a sequence of difference images formed from the input image sequence and the calculated background image sequence using a threshold, the sequence of the object masks representing shapes and positions of moving objects;

using said sequence of binary object masks for controlling spatial selectivity in a calculation of said background image sequence;

determining sizes and positions of centers of gravity of segments of said binary object masks; and

determining motion vectors of detected objects by minimization of squared gray scale value differences averaged over intersections of shifted mask segments.

2. The method according to claim 1, wherein noise suppression is carried out in the object masks by using median filters.

3. The method according to claim 1, wherein initial motion vectors, predicted by Kalman filters, are used for carrying out the minimization.

4. The method according to claim 1, wherein two or more images of the background are stored in a background memory, or one image of the background and a difference between two images of the background are stored in the background memory and wherein a recursion equation based on the stored images is used for calculating the images of the background, which produce the result that gradual brightness and contrast changes do not cause any deviation between a current image of the background and a stored image of the background.