mapping from camera/Image-raw to kinect RGB-D
Hi every body I'm working on follow-me task. The goal is to make a turtle-bot-like robot follow a specific person in the crowd. I use skeleton_markers to find joints of target person's body and with some control procedure follow her/him. To better support the gained information from skeleton_markers, I want to know how to find corresponding pixel(s) of a specific joint, so I will be able to extract some visual features to uniquely identify the followed person in the crowd.
Concisely, what is the x-y (in pixel) of a joint (that skeleton_markers has published) so I can draw a bounding-box around it. And also in case of losing the followed person (for example due to changing pose), I want to have an approximate place -based on the aforementioned bounding-box- to look for the person in that area.
Can somebody tell any package or method/procedure to achieve this? Any help would be greatly appreciated. P.S.: I'm using ROS Hydro on Ubuntu 12.04. P.P.S: Packages like OpenTLD can follow a textured, but I can't find a way to hint the skeleton_markers package to search a specific place to find the person in case of losing her/him.
