Posture Estimation System for Excavator Manipulator Using Deep Learning and Inverse Kinematics

Abstract

This study aims to construct a simple system for estimating and measuring the front manipulator posture by attaching a camera to an excavator. We propose a method to extract the feature points of the arms and buckets tip by deep learning and to calculate the posture of the front part using camera geometry and inverse kinematics from the obtained feature points. The effectiveness of the proposed method was verified experimentally using a scale model.