Fast Object Pose Estimation Using Adaptive Threshold for Bin-Picking

Wu Yan,Hongmin Wu,Zhihao Xu,Xuefeng Zhou,Shuai Li,Qianxing Su

doi:10.1109/access.2020.2983173

Abstract

Robotic bin-picking is a common process in modern manufacturing, logistics, and warehousing that aims to pick-up known or unknown objects with random poses out of a bin by using a robot-camera system. Rapid and accurate object pose estimation pipelines have become an escalating issue for robot picking in recent years. In this paper, a fast 6-DoF (degrees of freedom) pose estimation pipeline for random bin-picking is proposed in which the pipeline is capable of recognizing different types of objects in various cluttered scenarios and uses an adaptive threshold segment strategy to accelerate estimation and matching for the robot picking task. Particularly, our proposed method can be effectively trained with fewer samples by introducing the geometric properties of objects such as contour, normal distribution, and curvature. An experimental setup is designed with a Kinova 6-Dof robot and an Ensenso industrial 3D camera for evaluating our proposed methods with respect to four different objects. The results indicate that our proposed method achieves a 91.25% average success rate and a 0.265s average estimation time, which sufficiently demonstrates that our approach provides competitive results for fast objects pose estimation and can be applied to robotic random bin-picking tasks.

Highlights

An accurate, fast, and robust 6D object pose estimation solution has served an important role in many practical applications such as robot manipulation, augmented reality and autonomous driving
The RGB-D camera is a critical device for obtaining visual information from commercial to industrial levels, such as Kinect, RealSense, Bumblebee, etc., which has been proven in deriving better performances in 6D object pose estimation with respect to the RGB cameras without depth or distance measurement
We propose a fast CAD-based 6-DoF pose estimation pipeline for random bin-picking for different objects of varying shapes. 3D CAD data or target partial point clouds were used to create an off-line database that includes different feature information about the model

Summary

INTRODUCTION

Fast, and robust 6D object pose estimation solution has served an important role in many practical applications such as robot manipulation, augmented reality and autonomous driving. The former is textureless, and the later is thin and partially occluded. Concerning the pose estimation task, a voting-based feature matching scheme was introduced to perform comparison scheduling using the target in the database and real scenes to create pose clustering.

SYSTEM ARCHITECTURE

EXPERIMENTAL VERIFICATION

Findings

CONCLUSION AND FUTURE WORK