Method for detecting and separating the shadow of moving objects in a sequence of digital images

• US 5,592,567 A
• Filed: 05/10/1995
• Issued: 01/07/1997
• Est. Priority Date: 11/10/1992
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Kilger, Michael
• Primary Examiner(s): Couso, Jose L.
• Assistant Examiner(s): Tadayon, Bijan
• Application Number:
  US08/416,838
• Time in Patent Office: 608 Days
• Field of Search: 382/103, 382/170, 382/283, 382/107, 382/104, 382/168, 382/169, 382/171, 382/172, 382/199, 382/203, 348/169, 348/571, 348/586, 348/576, 348/416, 348/578, 348/579, 348/671, 348/672, 348/135, 348/148, 348/149, 395/126, 395/133, 395/134, 395/118, 395/120, 395/121, 395/135, 340/942, 340/937, 340/903, 340/435, 340/933, 340/938
• US Class Current: 382/199
• CPC Class Codes: G06T 7/246 using feature-based methods, e.g. the tracking of corners or segments
• CPC Class Codes: G06V 10/273 removing elements interfering with the pattern to be recognised
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats
• CPC Class Codes: G08G 1/015 with provision for distinguishing between two or more types of vehicles, e.g. between motor-cars and cycles
• CPC Class Codes: G08G 1/04 using optical or ultrasonic detectors

Abstract

For detecting and separating the shadow of moving objects in a sequence of digital images, a sequence of background images is first determined from a sequence of images, this sequence of background images containing only the non-moving image background of the sequence of images but not the moving objects. Object edges are then extracted from the sequence of images in that in each case an edge gray-value image is determined for each image of the sequence of images and for each respective corresponding background image of the sequence of background images. Using vertical or horizontal projection of these edge gray-value images, corresponding vertical or horizontal edge histograms are determined in each case for the edge gray-value images. Edges which do not belong to moving objects or shadows of moving objects are eliminated, in that the edge histograms of a background image are subtracted from the corresponding edge histograms of the image, belonging to the same instant, of the sequence of images and in that the differential edge histograms formed in this manner are subjected to a threshold value discrimination, by which means those edges are eliminated which do not correspond to the actual moving objects but rather to a shadow of a moving object.

Claims

1. A method for detecting and separating shadow of moving objects in a sequence of digital images, comprising the steps of:

a) determining a sequence of background images from the sequence of digital images, said sequence of background images containing only the non-moving image background of the sequence of images but not the moving objects;

b) determining, for extracting object edges, in each case an edge gray-value image for each image of the sequence of digital images and for each respective corresponding background image of the sequence of background images;

c) determining, by means of vertical or horizontal projection of said edge gray-value images, corresponding vertical or horizontal edge histograms in each case for the edge gray-value images;

d) eliminating edges which do not belong to moving objects or shadows, in that edge histograms of a background image are subtracted from corresponding edge histograms of an image, at a common instant in time, of the sequence of digital images, thereby forming differential edge histograms;

e) subjecting the differential edge histograms to a threshold value discrimination, to eliminate those edges which do not correspond to actual moving objects but which do correspond to a shadow of a moving object.

2. The method as claimed in claim 1, wherein the edge gray-value images are put into binary form before the determination of the edge histograms.

3. The method as claimed in claim 1, wherein, before carrying out steps b) to e) image regions are determined which correspond to moving objects and wherein steps b) to e) are applied only to these image regions.

4. The method as claimed in claim 1, wherein the method further comprises forming a time sequence of edge histograms from the edge histograms which belong to a common moving object, and improving elimination of shadow edges by identification and time tracking of corresponding edges.

5. The method as claimed in claim 1, wherein the method further comprises elimination of shadow edges by evaluating a preknowledge about a lateral position of the shadow edges with reference to corresponding object edges.

6. The method as claimed in claim 4, wherein the method further comprises elimination of shadow edges by evaluating a preknowledge about a lateral position of the shadow edges with reference to corresponding object edges, and wherein the preknowledge about the lateral position of the shadow edges is learned in an adaptive manner using the time sequence and is continuously checked.