How can I build a tf which is required by gampping?

edit

I finished base_link → odom tf like this

    map
     |
   odom
     |
 base_link

But in another required tf transform, it reads

(the frame attached to incoming scans) → base_link usually a fixed value, broadcast periodically by a robot_state_publisher, or a tf static_transform_publisher.

( http://wiki.ros.org/gmapping )

I can not understand how to build tf: scan→ base_link , and depthimagine_to_laserscan(I use kinect) do not provide any tf Transform.

And tf transforms in Turtlebot like this:

                                 map
                                  |
                                odom
                                  |
                              base_link
                                  |
                  ---------camera_rgb_frame-------
                  |               |              | 
    camera_depth_frame     camera_link   camera_rgb_optical_frame
                  |
    camera_depth_optical_frame

So, how can I combine scan data to base_link and broadcast this tf like turtlebot?

Some keywords like "camera_depth_frame", "camera_link"... Broadcasting them direct without any define or export?

And my kinect is fixed on the underpan, does it need a URDF file, or just use static_transform_publisher is fine?

Sincerely