Target Recognition and Localization-based Manipulator Control

Abstract

The ocean is rich in resources., but the harsh marine environment poses a great danger to manual fishing operations. Underwater robots have great potential in the field of underwater operations and can be equipped with robotic arms for fishing applications, thus improving the efficiency of fishing operations as well as reducing the operational risks. In this paper, we study a robotic arm control system based on the target recognition and target localization. We use YOLO v3 algorithm to identify sea cucumbers and the AP is 94.78% on the sea cucumber image test set. We calculate the 3D coordinates of the sea cucumber by binocular camera and convert the position of the sea cucumber to the coordinate system of the robot arm, and then we can control the movement of the end of the robot arm to the location of the sea cucumber.