Method and apparatus for tracking target object

1. A method for tracking a target object, comprising:

A. receiving a plurality of images, and selecting one of the images as a current image;

B. extracting a specific color of the current image;

C. comparing the current image with a template image to search a target object in the current image;

D. determining whether the current image contains the target object, and if yes, updating an object feature and an object location of the target object into a storage unit; and if not, performing a next step:

E. using a micro processor to perform the following steps:

a. searching a previous image with the target object in the images received before the current image;

b. searching whether the current image contains the target object according to the object feature of the previous image; and

c. finding the target object of the current image according to the object feature of the previous image, and updating the object feature and the object location of the current image into the storage unit.

2. The method for tracking a target object according to claim 1 , wherein the step B comprises:

respectively performing an interpolation on a vector converted from the specific color and vectors converted from a plurality of color pixels of the current image to respectively obtain a color similarity between each of the color pixels and the specific color; and

converting the current image into a gray-level image according to the color similarity, wherein the step is to respectively convert the color pixels into a gray-level intensity value according to the following formula:

I=w n ×V max , n≧ 1;

wherein w is the color similarity, V max is a maximum red-green-blue (RGB) value in the color pixels, I is the gray-level intensity value, and n is a coefficient for adjusting the color similarity.

3. The method for tracking a target object according to claim 1 , wherein the step C comprises:

performing an edge detection on the gray-level image to find an object edge;

respectively calculating a distance transform according to a plurality of angles of the object edge to obtain a distance transform image; and

performing a convolution operation on the template image and the distance transform image to obtain an image similarity between the template image and the distance transform image; and

the step D comprises:

comparing the image similarity with a first threshold to determine whether an object corresponding to the object edge is the target object;

if the image similarity is greater than the first threshold, determining that the object corresponding to the object edge is the target object; and

if the image similarity is not greater than the first threshold, determining that the object corresponding to the object edge is not the target object.

4. The method for tracking a target object according to claim 1 , further comprising:

specifying the specific color, by at least one of the following methods:

presetting the specific color;

specifying the specific color according to a selection on a displayed picture by a user; and

automatically searching the target object in the displayed picture through a self-learning mechanism, and extracting the color of the target object as the specific color.

5. The method for tracking a target object according to claim 1 , wherein the step b comprises:

b-1. establishing a feature block by using the object feature of the previous image;

b-2. searching a corresponding search range in the current image according to the object location of the previous image; and

b-3. searching whether the current image contains the target object within the search range by using the feature block.

6. The method for tracking a target object according to claim 5 , wherein the object feature comprises at least one of shape information, color information, depth information and movement track information of the target object.

7. The method for tracking a target object according to claim 5 , wherein the step b-3 comprises:

selecting a matching block within the search range;

calculating a pixel number of the matching block matching with the feature block; and

if the pixel number is greater than a second threshold, determining that the matching block is the target object.

8. The method for tracking a target object according to claim 1 , wherein after the step b, the method further comprises:

if the target object is not found in the current image according to the object feature of the previous image, determining that the current image does not contain the target object.

9. The method for tracking a target object according to claim 1 , further comprising:

using a first image capturing unit to capture the images;

calculating an epipolar line of a second image capturing unit according to the object location of the current image;

using the object feature of the current image to search the object location on the epipolar line of an image captured by the second image capturing unit; and

calculating three-dimensional coordinates of the target object according to the object location respectively obtained by the first image capturing unit and the second image capturing unit.

10. An apparatus for tracking a target object, comprising:

a storage unit, and

a micro processor, coupled to the storage unit, and programmed to perform the following steps:

A. receiving a plurality of images, and selecting one of the images as a current image;

B. extracting a specific color of the current image;

C. comparing the current image with a template image to search a target object in the current image;

D. determining whether the current image contains the target object, and if yes, updating an object feature and an object location of the target object into the storage unit; and if not, performing a next step:

E. using the micro processor to perform the following steps:

a. searching a previous image with the target object in the images received before the current image;

b. searching whether the current image contains the target object according to the object feature of the previous image; and

c. finding the target object of the current image according to the object feature of the previous image, and updating the object feature and the object location of the current image into the storage unit.

11. The apparatus for tracking a target object according to claim 10 , further comprising:

a first image capturing unit, coupled to the micro processor, for capturing the images; and

a second image capturing unit, coupled to the micro processor, for capturing another image at a different viewing angle;

wherein the micro processor calculates an epipolar line of the second image capturing unit according to the object location of the current image, uses the object feature of the current image to search the object location on the epipolar line of the image captured by the second image capturing unit, and calculates three-dimensional coordinates of the target object according to the object location respectively obtained by the first image capturing unit and the second image capturing unit.

12. The apparatus for tracking a target object according to claim 10 , wherein the step B performed by the micro processor comprises:

respectively performing an interpolation on a vector converted from the specific color and vectors converted from a plurality of color pixels of the current image to respectively obtain a color similarity between each of the color pixels and the specific color; and

converting the current image into a gray-level image according to the color similarity, wherein the step is to respectively convert the color pixels into a gray-level intensity value according to the following formula:

I=w n ×V max , n≧ 1;

wherein w is the color similarity, V max is a maximum red-green-blue (RGB) value in the color pixels, I is the gray-level intensity value, and n is a coefficient for adjusting the color similarity.

13. The apparatus for tracking a target object according to claim 10 , wherein the step C performed by the micro processor comprises: performing an edge detection on the gray-level image to find an object edge, respectively calculating a distance transform according to a plurality of angles of the object edge to obtain a distance transform image, and performing a convolution operation on the template image and the distance transform image to obtain an image similarity between the template image and the distance transform image; and

the step D performed by the micro processor comprises: comparing the image similarity with a first threshold to determine whether an object corresponding to the object edge is the target object; if the image similarity is greater than the first threshold, determining that the object corresponding to the object edge is the target object; and

if the image similarity is not greater than the first threshold, determining that the object corresponding to the object edge is not the target object.

14. The apparatus for tracking a target object according to claim 10 , wherein the micro processor specifies the specific color by at least one of the following methods:

presetting the specific color;

specifying the specific color according to a selection on a displayed picture by a user; and

automatically searching the target object in the displayed picture through a self-learning mechanism, and extracting the color of the target object as the specific color.

15. The apparatus for tracking a target object according to claim 10 , wherein the step b performed by the micro processor comprises: establishing a feature block by using the object feature of the previous image, and searching a corresponding search range in the current image according to the object location of the previous image; and selecting a matching block within the search range, calculating a pixel number of the matching block matching with the feature block, and if the pixel number is greater than a second threshold, determining that the matching block is the target object.

16. The apparatus for tracking a target object according to claim 10 , wherein the object feature comprises at least one of shape information, color information, depth information and movement track information of the target object.