Object identification method and device

1. An object identification method, comprising:

establishing a training database comprising a photographing distance of a training image and a training camera parameter;

obtaining a test image, a depth image, an RGB image, and a test camera parameter during the photographing of a target test object, wherein the RGB image includes the R/G/B gray scale value of each pixel of the target test object;

adjusting the RGB image according to the training database, the depth image, and the test camera parameter, such that the adjusted RGB image has a size equivalent to that of the training image in the training database;

wherein

the scaling of the test image is determined according to the photographing distance of the training image, a photographing distance of the target test object, the training camera parameter, and the test camera parameter;

a plurality of candidate regions comprising a plurality of object features are generated from the test image;

each candidate region is scaled individually with reference to the average distance of all objects within each candidate region, wherein each scaled candidate region has a size equivalent to that of the training image; and

object identification is performed on each scaled candidate region.

2. The object identification method according to claim 1 , wherein, the training camera parameter comprises any combinations of focal length parameter, magnification power, and camera resolution.

3. The object identification method according to claim 1 , further comprising:

dividing the scaled test image into a plurality of test sub-images each having a size equivalent to that of the training image; and

performing object identification on the test sub-images.

4. An object identification device, comprising:

a database module configured to establish a training database comprising a photographing distance of a training image and a training camera parameter;

a controller configured to obtain a test image, a depth image, an RGB image, and a test camera parameter during the photographing of a target test object, wherein the RGB image includes the R/G/B gray scale value of each pixel of the target test object; and

an identification module configured to adjust the RGB image according to the training database, the depth image, and the test camera parameter, the identification module, such that the adjusted RGB image has a size equivalent to that of the training image in the training database;

wherein, the controller is configured to

determine the scaling of the test image according to the photographing distance of the training image, a photographing distance of the target test object, the training camera parameter, and the test camera parameter;

generate a plurality of candidate regions comprising a plurality of object features from the test image;

scale each candidate region individually with reference to the average distance of all objects within each candidate region, wherein each scaled candidate region has a size equivalent to that of the training image; and

perform object identification on each scaled candidate region.

5. The object identification device according to claim 4 , wherein, the training camera parameter comprises any combinations of focal length parameter, magnification power, and camera resolution.

6. The object identification device according to claim 4 , wherein, the controller is configured to:

divide the scaled test image into a plurality of test sub-images each having a size equivalent to that of the training image; and

perform object identification on the test sub-images.