Optimization of target acquisition and sorting for object-finding multi-manipulator based on open MV vision

Abstract

To optimize the mechanical arm target capture and classification of the open multiple-view (MV) visualization program, the open MV visualization programming and deep learning detection method combined with the different capture strategies of robotic arm, a method to extend the research is proposed. For the proposed sorting robot’s multi-cargo grasping, the analysis required to detect a wide variety of goods in a storage environment that lacks color or structural features uniformly. On the basis of SSD target detection method regression, the object’s 3D position information is reconstructed by default preselected cell selection. 3D coordinate accuracy of binocular navigation system was verified as 8% when the target cargo location distance is more than 5 cm, and binoculars matching success rate is 89.7%. The success rate of Sorting and hoarding is increased from 6% to 85% by adding a change to the scoring points of the target products of uneven quality, with this we have achieved efficient and accurate import.

Keywords

• open mv vision
• target capture
• many mechanical arm