Object detection method and device

Patent metadata

Field	Value
Publication number	US-2016019683-A1
Application number	US-201514797667-A
Country	US
Kind code	A1
Filing date	Jul 13, 2015
Priority date	Jul 17, 2014
Publication date	Jan 21, 2016
Grant date	—

Abstract

Disclosed is an object detection method used to detect an object in an image pair corresponding to a current frame. The image pair includes an original image of the current frame and a disparity map of the same current frame. The original image of the current frame includes at least one of a grayscale image and a color image of the current frame. The object detection method comprises steps of obtaining a first detection object detected in the disparity map of the current frame; acquiring an original detection object detected in the original image of the current frame; correcting, based on the original detection object, the first detection object so as to obtain a second detection object; and outputting the second detection object.

First claim

What is claimed is:

1 . An object detection method used to detect an object in an image pair corresponding to a current frame, the image pair including an original image of the current frame and a disparity map of the same current frame, the original image of the current frame including at least one of a grayscale image and a color image of the current frame, the object detection method comprising:
an obtainment step of obtaining a first detection object detected in the disparity map of the current frame; an acquisition step of acquiring an original detection object detected in the original image of the current frame; a first correction step of correcting, based on the original detection object, the first detection object so as to obtain a second detection object; and an output step of outputting the second detection object.

2 . The object detection method according to claim 1 , wherein, the acquisition step includes:
obtaining a historical object template and plural historical weight coefficients for the historical object template, the historical object template being divided into plural historical segments which are respectively assigned the plural historical weight coefficients; determining, based on the historical object template, one or more original candidate objects in the original image of the current frame; respectively calculating, based on the plural historical segments and the plural historical weight coefficients, weighted similarities between the historical object template and the one or more original candidate objects; and determining an original candidate object, whose relevant weighted similarity is maximum, as the original detection object.

3 . The object detection method according to claim 2 , wherein, the respectively calculating, based on the plural historical segments and the plural historical weight coefficients, the weighted similarities between the historical object template and the one or more original candidate objects includes:
regarding each of the one or more original candidate objects,
dividing the corresponding original candidate object into plural original segments in a same way as the historical object template is divided into the plural historical segments;
respectively calculating plural segment similarities between the plural original segments and the plural historical segments;
respectively weighting the plural segment similarities by using the plural historical weight coefficients; and
calculating a sum of the plural weighted segment similarities so as to serve as the weighted similarity between the corresponding original candidate object and the historical object template.

4 . The object detection method according to claim 2 , further comprising:
a second correction step of correcting, based on the first detection object, the plural historical weight coefficients so as to obtain plural current weight coefficients which will serve as historical weight coefficients with respect to an original image of a next frame.

5 . The object detection method according to claim 4 , wherein, the second correction step includes:
respectively determining plural disparity point distributions of the first detection object in plural original segments obtained by dividing the original detection object in a same way as the historical object template is divided into the plural historical segments; and respectively generating, based on the plural disparity point distributions, the plural current weight coefficients for the plural original segments of the original detection object.

6 . The object detection method according to claim 1 , wherein, the obtainment step includes:
obtaining a historical detection object in a disparity map of a historical frame; based on the historical detection object, estimating, in the disparity map of the current frame, a current estimation object corresponding to the historical detection object; determining one or more first candidate objects in the disparity map of the current frame; respectively determining matching degrees between the current estimation object and the one or more first candidate objects; and determining a first candidate object, whose relevant matching degree is maximum, as the first detection object.

7 . The object detection method according to claim 6 , wherein, the acquisition step includes:
acquiring the original detection object detected in a predetermined range of the current estimation object in the original image of the current frame.

8 . The object detection method according to claim 1 , wherein, the first correction step includes:
determining, based on the original detection object, a current correction region in the disparity map of the current frame; and generating the second detection object based on at least a set of disparity points included in the current correction region.

9 . The object detection method according to claim 1 , before the first correction step, further comprising:
a determination step of determining whether the first detection object and the original detection object match each other.

10 . The object detection method according to claim 9 , wherein, the determination step includes at least one of steps of:
determining whether an overlap area between the first detection object and the original detection object is greater than or equal to a predetermined threshold, and if so, then determining that the first detection object and the original detection object match each other; and determining whether a distance between the first detection object and the original detection object is less than or equal to a predetermined threshold, and if so, then determining that the first detection object and the original detection object match each other.

11 . An object detection device used to detect an object in an image pair corresponding to a current frame, the image pair including an original image of the current frame and a disparity map of the same current frame, the original image of the current frame including at least one of a grayscale image and a color image of the current frame, the object detection device comprising:
an obtainment part configured to obtain a first detection object detected in the disparity map of the current frame; an acquisition part configured to acquire an original detection object detected in the original image of the current frame; a first correction part configured to correct, based on the original detection object, the first detection object so as to obtain a second detection object; and an output part configured to output the second detection object.

Assignees

• You Ganmei
• Liu Yuan
• Shi Zhongchao
• Lu Yaojie
• Wang Gang
• Ricoh Co Ltd

Inventors

• You Ganmei
• Liu Yuan
• Shi Zhongchao
• Lu Yaojie
• Wang Gang

Classifications

• G06T7/74 Primary

  involving reference images or patches · CPC title

• G06T2207/30261

  Obstacle · CPC title

• G06T2207/10021

  Stereoscopic video; Stereoscopic image sequence · CPC title

• G06T2207/10028

  Range image; Depth image; 3D point clouds · CPC title

• G06T2207/10024

  Color image · CPC title