how to recogize objects with kinect camera using RGB-D images?

I'm working now with object recognition. I have seen this video here and doing the car finder example using my webcam.

Now I want that my software recognizes the objects with kinect, not as normal images (RGB) but as RGB-D or 3D images, also color and depth. i.e. I need visual and depth informations of the images, then the both should be used by feature extraction.

Also, how to combine the depth and visual informations together to extract features?