Weiyang Zhao, Jianjun Yan, Houru Chen, Lianxin Gao, and Wei Zhou
Acupuncture points location, human posture estimation recognition,3D image, gradient boosting decision tree, TCM massage
We present an approach to automatically locate the human back acupuncture points based on RGB and depth images for traditional Chinese medicine (TCM) massage robots. The approach extracted key points on the human back to learn the position of acupuncture points. The key points around shoulders were extracted from RGB image using OpenPose network. A buttock key point was further defined and obtained from the depth image according to depth information of the buttocks. Then the key points were fed into the gradient boosting decision tree (GBDT) algorithm to learn association between human key points and acupuncture points. Compared to the majority of traditional location approach which still requires manual identification, our approach is fully automatic. What is more, the mean per acupuncture point position error of our approach is 10.57 mm, which meets the requirements of acupuncture points location in TCM massage robots and shows the potential of our system in terms of practical use.
Important Links:
Go Back
