120 likes | 269 Views
Machine Vision Applications. Handling Semi-flexible Objects: Scissors. Physical Layout. Overhead camera Scissors lie on a flat-topped table Scissors can be at any position and orientation but must lie entirely within the camera's field of view
E N D
Machine Vision Applications Handling Semi-flexible Objects: Scissors
Physical Layout • Overhead camera • Scissors lie on a flat-topped table • Scissors can be at any position and orientation but must lie entirely within the camera's field of view • Scissors can be fully closed, partially open, or fully open
Sample Images(N.B. Euler number changes: open, -1; closed, -2)
Guiding a Robot Using Vision Start with a binary image of a pair of scissors, which may opened by an arbitrary amount (fully closed to fully open). • The vision system locates and determines the orientation of the larger of the two finger hole. • The robot places its gripper in the larger finger hole & picks up the scissors. • The robot closes the scissors by pressing one of the 2 "free" ends against an upright pillar. • The robot packs the scissors in an appropriately shaped slot in a tray of tools. • Using another camera, the vision system checks the final placement of the scissors. N.B. The vision system and the robot act together. This procedure works whatever the opening angle.
Determining whether the scissors are fully closed Use Euler number (eul in QT): Equal to -1 if scissors are open Equal to -2 if they fully closed. N.B. The largest lake is always the larger finger hole, whatever the opening angle.
Obtaining a Secure Grasp • Locate the finger holes as these are invariate. • Since we eventually need the scissors closed, we can place a gripper in the larger finger hole. • Then, lift the scissors and press them against a bar, or post, to push them closed. • The robot can "normalise" the scissors (i.e. close them). This is easier than relying on the vision system to compensate for the variable shape.
Locating the Larger Finger Hole (i.e. largest lake) QT command sequence: blf % Fill lakes exr % Exclusive OR big % Retain only the biggest blob [x,y] = cgr; % Position of blob centroid z = lmi; % Orientation
Chirality • Normalise the scissors (i.e close them) • Draw a straight line from the centroid of larger finger holes to the centroid of the smaller finger hole (Use big(2) to isolate this.) • Find whether the centroid of the scissors is to the left or right of this line. This defines the chirality arbitrarily but consistently.
Estimating the degree of opening Measure the distance between the centroids of the finger holes (lakes). This increases monotonically as the scissors are opened wider. When the blades are nearly closed, locating the tips is difficult; when they are fully closed, it is impossible. Locating the centroids of the finger holes is easier and more reliable than locating the tips of the blades.
QT Command Sequence: [x0,y0] = cgr; % Centroid of the scissors blf % Fill lakes exr % Exclusive OR big(1) % Retain only the biggest blob [x1,y1] = cgr; % Position of centroid of largest blob swi % Interchange images big(2) % Retain only the second biggest blob [x2,y2] = cgr; % Centroid of 2nd largest blob s = (x0-x1)*(y2-y1) -(y0-y1)*(x2-x1) % s = 0 corresponds to straight line % from [x1,y1] to [x2,y2] chirality = sign(s) % Is s > 0?