Research And Design Of Automatic Moxibustion Device Forhuman Hand

Acupuncture point therapy is being practiced widely in different parts of the world.Acupoint stimulation has proven to be of significant importance for rehabilitation and preventive therapy.Moxibustion,a kind of acupoint therapy,has mainly been performed by practitioners relying on manual localization and positioning of acupoints,leading to variance in the accuracy owing to human error.Developments in the automatic detection of acupoints using electronic devices have proven to assist the practitioners in locating the acupoint accurately.But there remains a great need to make the localization of acupoints completely automatic.Computer vision techniques have been deployed in the past to detect acupoints of different body parts.Deep learning techniques have proven to somewhat tackle the problem of autonomous detection using image processing and neural networks.But the current methods lack depthbased localization and are thus confined to two-dimensional(2D)localization.In order to develop a system that can detect accurate positioning of the acupoints and perform robotcontrolled stimulation,the 3D coordinate information of the acupoints must be considered.In this research,a novel way of 3D acupoint localization is introduced,based on a fusion of RGBCNN and depth-CNN to extract the features and landmarks of human hand thus calculating the acupoint’s 3D coordinate information and guiding the manipulator.The overall research accomplishments are as follows:Firstly,a 3D visual sensor(Kinect v1)is calibrated using a checkerboard pattern and intrinsic and extrinsic parameters are obtained from the calibration process,then the transformation matrix is computed to project the depth data into the RGB domain.Secondly,a 3D acupuncture point hand dataset is constructed for localization and visualization of the acupuncture points.Also,the RGB-D data obtained from the Kinect Vl sensor is converted to point cloud representation.The point cloud data is compressed and the valid points are extracted from the input overall point cloud,in order to remove the outliers that do not belong to the surface of the object owing to the different sensor limitations,object reflectiveness and different light intensity levels.Thirdly,a fusion of RGB-CNN and depth-CNN is employed,in order to obtain 3D localization.Then,3D coordinates are fed to the manipulator to perform artificially controlled moxibustion therapy.Furthermore,a 3D acupoint dataset consisting of RGB and depth images of hands,is constructed to train,validate and test the network.Lastly,a 3D CAD design of the temperature controlled moxibustion using a multi-axis robot to perform the therapy is also presented.The Convolutional neural network(CNN)based localization is incorporated into the software part,present in the associated computer,which is wirelessly connected to a standalone control panel responsible for the operation of the system.An Ultrasonic sensor makes sure the temperature during the therapy stays in the prescribed limit and a non-contact temperature sensor controls the optimal distance between the moxa stick and the human skin.The network was able to localize 5 sets of acupoints with an average localization error of less than 0.09.Further experiments prove the efficacy of the approach and lay grounds for development of automatic moxibustion robots.