Human image tracking system, and human image detection and human image tracking methods thereof

Patent metadata

Field	Value
Publication number	US-9317765-B2
Application number	US-201414146432-A
Country	US
Kind code	B2
Filing date	Jan 2, 2014
Priority date	Aug 15, 2013
Publication date	Apr 19, 2016
Grant date	Apr 19, 2016

Abstract

A human image detection and tracking systems and methods are disclosed. A human image detection method comprises receiving a depth image data from a depth image sensor by an image processing unit, removing a background image of the depth image sensor and outputting a foreground image by the image processing unit, receiving the foreground image and operating a graph-based segment on the foreground image to obtain a plurality of graph blocks by a human image detection unit, determining whether a potential human region exists in the graph blocks, determining whether the potential human region is a potential human head region, determining whether the potential human head region is a real human head region, and regarding the position of the real human head region is the human image position by the human image detection unit if the potential human head region is the real human head region.

First claim

What is claimed is:

1. A human image detection method for detecting a human image position in a detection area, comprising:
receiving by an image processing unit a depth image data from a depth image sensor;
removing by the image processing unit a background image of the depth image data for obtaining a foreground image;
receiving the foreground image by a human image detection unit, which performs a graph-based segment process on the foreground image to obtain a plurality of graph blocks;
determining by the human image detection unit whether a potential human region exists in the graph blocks according to an area size of each graph block; determining by the human image detection unit whether the potential human region is a potential human head region according to a similarity between the potential human region and a hemisphere model if the potential human region exists;
determining by the human image detection unit whether the potential human head region is a real human head region according to an area size of a surrounding region adjacent to the potential human head region if the potential human region is the potential human head region; and
determining by the human image detection unit the position of the real human head region as the human image position if the potential human head region is the real human head region.

2. The method according to claim 1 , wherein the removing the background image of the depth image data comprises: removing a stationary object in the depth image data.

3. The method according to claim 2 , wherein the performing the graph-based segment process on the foreground image comprises: calculating a plurality of depth differences between two adjacent pixels of all pixels of the depth image data; and sorting out pixels having depth differences less than a first predetermined critical value from all pixels and classifying them as belonging to the same graph block.

4. The method according to claim 1 , wherein the determining whether a potential human region exists in the graph blocks comprises: calculating the area size of each graph block; and determining whether the area size of each graph block is greater than a second predetermined critical value.

5. The method according to claim 4 , wherein if the area size of at least one graph block is greater than the second predetermined critical value, the human image detection unit determines that the potential human region exists in the graph blocks.

6. The method according to claim 4 , wherein if none of the area sizes of the graph blocks is greater than the second predetermined critical value, the human image detection unit determines that the potential human region does not exist in the graph blocks, and the image processing unit receives another depth image data from the depth image sensor.

7. The method according to claim 1 , wherein the determining whether the potential human region is the potential human head region comprises: calculating a difference between a plurality of three dimensional pixel coordinates of the potential human head region and the hemisphere model, wherein a smaller difference represents a higher similarity, and a greater difference represents a lower similarity; and determining whether the similarity is greater than a third predetermined critical value.

8. The method according to claim 7 , wherein if the similarity is greater than the third predetermined critical value, the human image detection unit determines that the potential human region is the potential human head region.

9. The method according to claim 7 , wherein if the similarity is less than the third predetermined critical value, the human image detection unit determines that the potential human region is not the potential human head region, and the image processing unit receives another depth image data from the depth image sensor.

10. The method according to claim 1 , wherein the determining whether the potential human head region is the real human head region comprises: regarding a central position of the hemisphere model as the position of the potential human head region if the similarity is high; calculating a plurality of depth differences between a depth of the position of the potential human head region and a plurality of depths of pixels in the surrounding region adjacent to the potential human head region; adding sums of the depth differences, in which each depth difference is less than a fourth predetermined critical value; and determining whether the area of the surrounding region adjacent to the potential human head region is greater than a fifth predetermined critical value according to the sums of the depth differences.

11. The method according to claim 10 , wherein if the area of the surrounding region adjacent to the potential human head region is greater than the fifth predetermined critical value, the human image detection unit determines that the potential human head region is a reliable real human head region.

12. The method according to claim 10 , wherein if the area of the surrounding region adjacent to the potential human head region is less than the fifth predetermined critical value, the human image detection unit determines that the potential human head region is not the real human head region, and the image processing unit receives another depth image data from the depth image sensor.

13. A human image tracking method for tracking a human image in a detection area operated by a human image tracking unit connected to an image processing unit and a human image detection unit, the method comprising:
receiving a current human image position from the human image detection unit and a prior tracking position detected by the human image tracking unit;
determining whether a distance between the current human image position and the prior tracking position is less than a first predetermined critical value;
generating a plurality of supposed current human image positions adjacent to the prior tracking position if the distance is less than the first predetermined critical value;
calculating a plurality of depth distribution similarities between the supposed current human image positions and the prior tracking position;
calculating a plurality of accurate possibilities of the supposed current human image positions according to the plurality of depth distribution similarities; and
selecting the supposed current human image position corresponding to a maximum accurate possibility as a current tracking position of the human image.

14. The method according to claim 13 , wherein the human image tracking unit regards the current human image position as a latest tracking position if the distance is greater than the first predetermined critical value.

15. The method according to claim 13 , further comprising: the human image tracking unit determining whether the current tracking position is reliable based on a comparison of the maximum accurate possibility and a second predetermined critical value.

16. The method according to claim 15 , wherein if the maximum accurate possibility is greater than the second predetermined critical value, the human image tracking unit determines that the current tracking position is reliable and regards the current tracking position as a latest tracking position.

17. The method according to claim 15 , wherein if the maximum accurate possibility is less than the second predetermined critical value, the human image tracking unit determines that the current tracking position is unreliable and determines whether another current human image position is adjacent to the current tracking posit

Assignees

• Univ Nat Taiwan

Inventors

• Fu Li-Chen
• Tseng Ting-En
• Liu An-Sheng
• Hsiao Po-Hao

Classifications

• G06V40/10 Primary

  Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands · CPC title

• G06V10/255

  Detecting or recognising potential candidate objects based on visual cues, e.g. shapes · CPC title

• G06K9/00362

  Physics · mapped topic

• G06K9/3241 Primary

  Physics · mapped topic