Detection of objects in digital images

• US 8,509,526 B2
• Filed: 04/13/2011
• Issued: 08/13/2013
• Est. Priority Date: 04/13/2010
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Haas, Norman, Li, Ying, Pankanti, Sharathchandra
• Primary Examiner(s): Bali, Vikkram
• Application Number:
  US13/086,023
• Publication Number:
  US 20110249867A1
• Time in Patent Office: 853 Days
• Field of Search: None
• US Class Current: 382/162
• CPC Class Codes: G06F 18/2148 characterised by the process organisation or structure, e.g. boosting cascade
• CPC Class Codes: G06V 10/7747 Organisation of the process, e.g. bagging or boosting
• CPC Class Codes: G06V 20/582 of traffic signs

Abstract

A system and method to detect objects in a digital image. At least one image representing at least one frame of a video sequence is received. A given color channel of the image is extracted. At least one blob that stands out from a background of the given color channel is identified. One or more features are extracted from the blob. The one or more features are provided to a plurality of pre-learned object models each including a set of pre-defined features associated with a pre-defined blob type. The one or more features are compared to the set of pre-defined features. The blob is determined to be of a type that substantially matches a pre-defined blob type associated with one of the pre-learned object models. At least a location of an object is visually indicated within the image that corresponds to the blob.

Claims

1. An information processing system for detecting objects in a digital image, the information processing system comprising:

a memory;

a processor communicatively coupled to the memory; and

an object detection system communicatively coupled to the memory and the processor, the object detection system configured to perform a method comprising;

receiving at least one image representing at least one frame of a video sequence of an external environment;

extracting a given color channel of the at least one image;

identifying at least one blob that stands out from a background of the given color channel that has been extracted;

extracting one or more features from the at least one blob that has been identified;

providing the one or more features that have been extracted to a plurality of pre-learned object models each comprising a set of pre-defined features associated with a pre-defined blob type;

comparing, in response to the providing, the one or more features to the set of pre-defined features of each of the plurality of pre-learned object models;

determining, based on the comparing, that the at least one blob is of a type that substantially matches a pre-defined blob type associated with one of the pre-learned object models; and

visually indicating, in response to the determining, at least a location of an object within the at least one image that corresponds to the at least one blob.

2. The information processing system of claim 1, wherein the method performed by the information processing system further comprises:

applying a bounding box around the at least one blob;

analyzing two or more corner regions of the bounding box; and

determining a rough shape of the at least one blob based on the two or more corner regions comprising one of a portion of background and a potential portion of the at least one blob.

3. The information processing system of claim 1, wherein the method performed by the information processing system further comprises:

determining that a classification probability of the at least one blob with respect to the plurality of pre-learned object models is below a given threshold;

comparing the at least one blob with one or more blobs from at least one preceding image that has been assigned a blob type from the plurality of pre-learned object models;

determining that the at least one blob substantially matches with at least one of the one or more blobs; and

assigning the at least one blob to the blob type that has been assigned to the at least one of the one or more blobs in the at least one preceding image.

4. The information processing system of claim 1, wherein the method performed by the information processing system further comprises:

tracking the object for at least one or more subsequent images, wherein the tracking comprises;

identifying a search area S in the one or more subsequent images; and

localizing the object within S.

5. The information processing system of claim 4, wherein identifying the search area S comprises:

determining a size of S based on a size of the object within the at least one image, and a moving direction of the object, wherein the moving direction is based on a displacement of the object in an x-direction and a y-direction in a preceding image of the at least one image, wherein the size of the search area S is further determined based on a size change of the object from a preceding image to the at least one image and a size of the object in the at least one image.

6. A computer program product for detecting objects in a digital image, the computer program product comprising:

a non-transitory storage medium readable by a processing circuit and storing instructions for execution by the processing circuit for performing a method comprising;

receiving at least one image representing at least one frame of a video sequence of an external environment;

extracting a given color channel of the at least one image;

identifying at least one blob that stands out from a background of the given color channel that has been extracted;

extracting one or more features from the at least one blob that has been identified;

providing the one or more features that have been extracted to a plurality of pre-learned object models each comprising a set of pre-defined features associated with a pre-defined blob type;

comparing, in response to the providing, the one or more features to the set of pre-defined features of each pre-learned object model;

determining, based on the comparing, that the at least one blob is of a type that substantially matches a pre-defined blob type associated with one of the pre-learned object models; and

visually indicating, in response to the determining, at least a location of an object within the at least one image that corresponds to the at least one blob.

7. The computer program product of claim 6, wherein the instructions are further for:

applying a bounding box around the at least one blob;

analyzing two or more corner regions of the bounding box; and

determining a rough shape of the at least one blob based on the two or more corner regions comprising one of a portion of background and a potential portion of the at least one blob.

8. The computer program product of claim 6, wherein the instructions are further for:

determining that a classification probability of the at least one blob with respect to the plurality of pre-learned object models is below a given threshold;

comparing the at least one blob with one or more blobs from at least one preceding image that has been assigned a blob type from the plurality of pre-learned object models;

determining that the at least one blob substantially matches with at least one of the one or more blobs; and

assigning the at least one blob to the blob type that has been assigned to the at least one of the one or more blobs in the at least one preceding image.

9. The computer program product of claim 6, wherein the instructions are further for:

tracking the object for at least one or more subsequent images, wherein the tracking comprises;

identifying a search area S in the one or more subsequent images; and

localizing the object within S.

10. The computer program product of claim 8, wherein identifying the search area S comprises:

determining a size of S based on a size of the object within the at least one image, and a moving direction of the object, wherein the moving direction is based on a displacement of the object in an x-direction and a y-direction in a preceding image of the at least one image, wherein the size of the search area S is further determined based on a size change of the object from a preceding image to the at least one image and a size of the object in the at least one image.