Workpiece detection device and workpiece detection method

A workpiece detection method comprising:

• photographing one or more workpieces loaded on a mounting place with a monocular camera, an uppermost workpiece of the one or more workpieces being a target workpiece that is held by a gripper of a workpiece holding robot, and the workpiece holding robot including a surface detection sensor;
• performing trapezoid correction on photographed image data obtained by photographing the one or more workpieces with the monocular camera;
• selecting a model having a size and an angle with a highest degree of matching by carrying out a first pattern matching between an image of the trapezoid-corrected photographed image data and a first plurality of models that show images of the target workpiece and have a plurality of sizes and a plurality of angles, each model showing a planar shape of the target workpiece viewed from above, the plurality of sizes being different in a stepwise manner from a smallest size corresponding to a size of the planar shape of the target workpiece in a state in which a single workpiece is arranged on the mounting place to a largest size corresponding to a size of the planar shape of the target workpiece in a state in which a maximum number of workpieces are loaded on the mounting place, and the plurality of angles being angles at which each size of the model is rotated in a plane containing the planar shape;
• detecting a position and an angle of the target workpiece from among the loaded workpieces based on the model selected by the first pattern matching;
• calculating an actual load height of the target workpiece based on a hand height detected by the workpiece holding robot when the gripper is positioned above the target workpiece and the surface detection sensor detects a surface of the target workpiece, based on the detected position and angle of the target workpiece;
• selecting a model having a size and an angle with a highest degree of matching by carrying out a second pattern matching between the image of the trapezoid-corrected photographed image data and a second plurality of models that are either selected from among the first plurality of models based on the actual load height or newly created based on the actual load height, the second plurality of models being less than the first plurality of models;
• and
• re-detecting the position and the angle of the target workpiece based on the model selected by the second pattern matching.

A control device for controlling a robot control device comprising:

• a monocular camera configured to photograph one or more workpieces loaded on a mounting place, an uppermost workpiece of the one or more workpieces being a target workpiece that is held by a gripper of a workpiece holding robot, and the workpiece holding robot including a surface detection sensor;
• and
• a computer device connected to the monocular camera, wherein
• the computer device is configured
• to perform trapezoid correction on photographed image data obtained by photographing the loaded one or more workpieces with the monocular camera;
• to carry out a first pattern matching between an image of the trapezoid-corrected photographed image data and a first plurality of models that show images of the target workpiece and have a plurality of sizes and a plurality of angles so as to select a model having a size and an angle with a highest degree of matching, each model showing a planar shape of the target workpiece viewed from above, the plurality of sizes being different in a stepwise manner from a smallest size corresponding to a size of the planar shape of the target workpiece in a state in which a single workpiece is arranged on the mounting place to a largest size corresponding to a size of the planar shape of the target workpiece in a state in which a maximum number of workpieces are loaded on the mounting place, and the plurality of angles being angles at which each size of the model is rotated in a plane containing the planar shape;
• to perform a primary detection to detect a position and an angle of the target workpiece from among the loaded workpieces based on the model selected by the first pattern matching;
• to calculate an actual load height of the target workpiece based on a hand height detected by the workpiece holding robot when the gripper is positioned above the target workpiece and the surface detection sensor detects a surface of the target workpiece, based on the position and the angle of the target workpiece detected by the primary detection;
• and
• to perform a secondary detection to re-detect the position and the angle of the target workpiece based on a model having a size and an angle with a highest degree of matching by carrying out a second pattern matching between the image of the trapezoid-corrected photographed image data and a second plurality of models that are either selected from among the first plurality of models based on the actual load height or newly created based on the actual load height, the second plurality of models being less than the first plurality of models.