Accelerated object recognition in an image

• US 9,495,612 B2
• Filed: 11/25/2014
• Issued: 11/15/2016
• Est. Priority Date: 11/28/2013
• Status: Active Grant

• Current Assignee: Robert Bosch GmbH
• Original Assignee: Robert Bosch GmbH
• Inventors: Cebron, Nicolas, Yu, Jie
• Primary Examiner(s): Liew, Alex
• Application Number:
  US14/553,412
• Publication Number:
  US 20150146927A1
• Time in Patent Office: 721 Days
• Field of Search: 382224-228
• US Class Current: 1/1
• CPC Class Codes: G06V 30/2504 Coarse or fine approaches, e.g. resolution of ambiguities or multiscale approaches

Abstract

A method for recognizing an object (120) in an image (100), in which the recognition of the object (120) comprises a first scaling stage of a scaling region of the image (100), and at least a further scaling stage of the scaling region of the image (100); and in which the image (100) is subdivided into at least one image segment (110); a first decision being taken for the at least one image segment (110) on the first scaling stage of the scaling region as to whether the at least one image segment (110) is considered on the at least one further scaling stage of the scaling region for the recognition of the object (120).

Claims

1. A method for recognizing an object in an image generated by a video monitoring system, the video monitoring system including a processor, the method comprising:

providing, in the processor, a first scaling stage of a scaling region of the image, and at least one further scaling stage of the scaling region of the image;

subdividing the image, with the processor, into at least one image segment;

making, with the processor, a first decision for the at least one image segment on the first scaling stage of the scaling region as to determine whether or not the at least one image segment should be further scaled by scaling stage of the scaling region; and

before the first decision, a second decision is taken for the at least one image segment on the first scaling stage of the scaling region as to whether the at least one image segment is considered for the first decision.

2. The method according to claim 1, in which the first decision is based on weighted first individual decisions concerning first parameters which are relevant to the recognition of the object.

3. The method according to claim 2, in which each one of the first individual decisions results from investigating whether a value of a first parameter is below or above a first threshold value and/or inside or outside of an interval around the first threshold value.

4. A method for recognizing an object in an image generated by a video monitoring system, the video monitoring system including a processor, the method comprising:

providing, in the processor, a first scaling stage of a scaling region of the image, and at least one further scaling stage of the scaling region of the image;

subdividing the image, with the processor, into at least one image segment;

making, with the processor, a first decision for the at least one image segment on the first scaling stage of the scaling region as to whether the at least one image segment is considered on the at least one further scaling state of the scaling region; and

before the first decision, a second decision is taken for the at least one image segment on the first scaling stage of the scaling region as to whether the at least one image segment is considered for the first decision,

wherein the second decision results from second individual decisions concerning second parameters which are relevant to the recognition of the object, the second individual decisions being taken in a sequence ordered by rising complexity of the second parameters and building up one upon another until a termination condition is reached.

5. The method according to claim 4, in which the at least one image segment is considered for the first decision when the number of the second individual decisions taken up to the termination condition is below a second threshold value.

6. A method for recognizing an object in an image generated by a video monitoring system, the video monitoring system including a processor, the method comprising:

providing, in the processor, a first scaling stage of a scaling region of the image, and at least one further scaling stage of the scaling region of the image;

subdividing the image, with the processor, into at least one image segment;

making, with the processor, a first decision for the at least one image segment on the first scaling stage of the scaling region based on weighted first individual decisions concerning first parameters which are relevant to recognition of the object, and the first decision resulting in whether the at least one image segment is considered on the at least one further scaling stage of the scaling region; and

before the first decision, making a second decision for the at least one image segment on the first scaling stage of the scaling region as to whether the at least one image segment is considered for the first decision.

7. The method according to claim 6, wherein the second decision results from second individual decisions concerning second parameters which are relevant to the recognition of the object, the second individual decisions being taken in a sequence ordered by rising complexity of the second parameters and building up one upon another until a termination condition is reached.