System and method for optimizing tracker system

A method for optimizing a tracker system, the method comprising the steps of:

• acquiring a mesh data representing possible positions and orientations of an active marker on a tracked object, and a pose data representing possible poses of the tracked object under working conditions and at least one camera position and orientation, and establishing a coordinate system for the tracked object (101);
• computing a visibility value of each node on a mesh from all camera viewports for every pose of the pose data and generating a visibility value list representing each node, wherein the visibility value represents how many times a node is visible from at least one camera viewport, considering all pose data, and wherein the visibility from any camera per pose is counted once (102) and wherein if LED fiducials are used, the visibility computation is based on the LED's normal with respect to the object coordinate system, LEDs illumination cone angle and the known pose of a head;
• selecting the node with highest visibility value as an active marker placement node (103);
• removing nodes which are closer to the selected node than a non-maxima suppression threshold (104);
• calculating the total number of markers (selected and placed so far) that are visible for each pose to inspect which poses exceed a K marker limit, and if a pose satisfies this condition, the pose should be excluded from the pose database;
• (104-A)
• removing the pose(s) that meet the requirements of step 104-A (105);
• testing if a predetermined percentage of all poses have the predetermined number of selected nodes (106);
• repeating step (102) to step (106) if the test in step (106) is negative (106a);
• projecting the final selected node positions on an image plane of each camera viewport for each pose, and adding a predetermined amount of noise into 2D marker images and using the 2D marker images for calculating the pose of the mesh with respect to the camera viewports using a tracker algorithm (107);
• calculating pose error by comparing a plurality of estimated positions of the markers with a plurality of real positions of the markers and calculating pose coverage which is a percentage of poses that can be estimated with the tracker algorithm by comparing tracker algorithm results with the ground truth data (108);
• recording and outputting the pose error and coverage results together with a number and positions of the active marker placement nodes (109);
• testing if a new optimization parameter set, pose estimation parameter set, camera location set or 3D object mesh exist (110);
• starting over with new data from the step (101) if the test in step (110) is positive (110a);
• selecting a camera location set, a 3D mesh data, a marker tree, an optimization parameter set and a pose estimation algorithm parameter set which satisfy at least one constraint wherein at least one constraint is selected from a group consisting of choosing a parameter set with the minimum pose error, a system with the minimum hidden poses and a system with the minimum marker number (111).