Development of a vision system for grasping of micro-objects

Abstract

In this report, we propose a vision system for robot grasping in complex environments. Achieving an accurate grasp of a target object is dependent on the vision system and a certain tracking ability. Object detection methods are proposed to determine sharp edges on different types of objects. Hence this report studies and evaluates various methods to correctly detect sharp edges of objects such as edge and object detection algorithms using OpenCV as well as state-of-the-art Artificial Intelligence algorithms using YOLOV4. Implementing such algorithms require data preparation of different datasets and training of the Model used in YOLOV4 to achieve accurate detection of sharp edges on objects.