Pose Estimation of Six-axis Industrial Robots Based on Deep Learning

Abstract

Industrial robot pose estimation has important applications in industrial robot safety detection and abnormal pose analysis. Aiming at the current problems of low accuracy of industrial robot pose estimation, this paper proposes a pose estimation method of six-axis industrial robots based on deep learning. Firstly, this paper uses the object detection method to detect the six joint axes of the industrial robot, and then the detected results are converted into the pose of the industrial robot. Additionally, this paper constructs an image dataset for industrial robot pose. Our method lays the foundation for industrial robot safety detection and abnormal pose analysis. The experimental results on the self-constructed dataset show that our method can accurately estimate the pose of the industrial robot, and the mean average precision (mAP) is 82.98%. It also outperforms previous industrial robot pose estimation methods by a significant margin of 29.98% performance gain.