Algorithm for Tool Grasp Detection

Abstract

The following paper presents the development of an algorithm for the detection of the grip point for an object using gripper. The algorithm receives as input the segmentation of the desired tool, where background becomes white and the objects, black. Additionally it allows the user to define the dimensions, in pixels, of the gripper that will make the grasp, the level of admissible thickness of the tool to determine the grip points, in order to prevent the fit between the gripper and the object section being too loose or too tight, and the maximum degree of inclination between the object section to be grasp, and the surface of the gripper. It was possible to develop an algorithm that finds different points of grip on a tool and selects the best one based on a series of criteria that determine the characteristics that the point must fulfill to be considered the best grip, and finally, the grip points found in the same tool are compared for different search parameters entered by the user, in order to find those that allow a reduction in the execution time of the algorithm without reducing the quality of the grip.