Only released in EOL distros:  


The portrait_painter stage is the last stage of the portrait_bot project. It is responsible for converting the images created in face_contour_detector to instructions for the robots arm. For this we use the ee_cart_imped controller. Furthermore portrait_painter is responsible for defining the canvas the image will be drawn on.

The package consists of three nodes:


portrait_painter is part of the portrait_bot stack. To learn how it is used please refer to the portrait_bot/Tutorials/Running the portrait bot Tutorial.

How does it work?


LineExtractor receives the contour images from contour_detector. It uses potrace to create a plot file representing the contours of the edges in the image. Then it converts that plot file into a list of lines, whereat each line consists of an absolute start point and following relative points.

Since potrace returns contours / polygons each line will result in at least two parallel edges. LineExtractor tries to revert that effect, so only one line will be drawn. Afterwards it calls for each line the job_submission service of the PainterMain node. After the last line is drawn the Gripper is moved away from the picture to a position right of the head.


PainterMain provides the services canvas_setting and job_submission. It holds three Vector3 objects representing the upper left, upper right and lower left corner of the canvas. These Vector3 objects have to be set by ManualCanvasSetting using the canvas_setting service call before the robot can draw.

PainterMain accepts jobs from LineExtractor using the job_submission service. Each job represents one line of the image as a list of 2D-Points with values between 0 and 1. Using the 3D coordinates of the canvas these points are translated into points in 3D space.

PainterMain creates a Trajectory object - as described ee_cart_imped_action/Tutorials/Writing a Stiffness Controller - that navigates the arm to a point 10 cm above the first point of the line, to the actual point and then to every other point of the line.


Before you can begin drawing, the robot must know where in the 3D space the canvas is. To do this we have chosen that the user must move the arm to the upper left, upper right and lower left corner of the canvas manually. There is a Set Corners button in the gui that will call a alubsc/node_instr/manual_canvas_setting server provided by ManualCanvasSetting. After the button is clicked the GUI will display instructions like "move Gripper to upper left corner of the canvas and press button".

Once the last point is set the node will call the canvas_setting server of PainterMain to report the canvas coordinates.

Wiki: portrait_painter (last edited 2011-05-05 14:34:40 by Felix Endres)