Abstract: A robotic system includes a robot, a three-dimensional (3D) point cloud camera, and a controller. The camera outputs an image data file having two-dimensional (2D) color data and 3D depth data. The controller receives input criteria and the image data file, and detects 2D and 3D line segments in the image data file. Sets of new 3D lines are generated via back-projecting some of the 2D line segments onto some of the 3D line segments. The new 3D lines are grouped into intersecting lines, and the controller performs a normal vector grouping function on the intersecting lines to identify candidate planes. Intersecting planes are identified from the candidate planes, and candidate boxes are identified from the intersecting planes. The controller locates a target box from among the identified candidate boxes via the input criteria. A control action may be commanded on the target box by the end effector.

Abstract: A robotic system includes a robot, a three-dimensional (3D) point cloud camera, and a controller. The camera outputs an image data file having two-dimensional (2D) color data and 3D depth data. The controller receives input criteria and the image data file, and detects 2D and 3D line segments in the image data file. Sets of new 3D lines are generated via back-projecting some of the 2D line segments onto some of the 3D line segments. The new 3D lines are grouped into intersecting lines, and the controller performs a normal vector grouping function on the intersecting lines to identify candidate planes. Intersecting planes are identified from the candidate planes, and candidate boxes are identified from the intersecting planes. The controller locates a target box from among the identified candidate boxes via the input criteria. A control action may be commanded on the target box by the end effector.