Research On Head Reconstruction And Target Tracking Of TMS Robot

Transcranial magnetic stimulation(TMS)is a non-invasive treatment of encephalopathy which is widely used in clinic.It has the characteristics of noninvasive,low cost,simple operation and so on.However,there are some shortcomings in traditional TMS treatment.First of all,doctors can not accurately locate the treatment target for the first time based on personal experience when determining the treatment target of the patient’s head.Secondly,in the process of treatment,the patient’s head will move involuntarily,which will lead to the deviation of transcranial magnetic stimulation coil from the therapeutic target,which can not achieve effective target tracking and positioning.In order to solve the two limitations mentioned above,this thesis designs a TMS treatment robot system based on 3D reconstruction of head model and pose detection.In this thesis,the 3D model of the patient’s head is reconstructed based on the positioning cap and motion recovery structure algorithm.In the off-line treatment stage,the camera is used to collect the multi view images of the patient’s head wearing the positioning cap,and the SFM algorithm is used to reconstruct the three-dimensional model of the patient’s head.With the help of the precise mark points of the positioning cap in the reconstruction model,the doctor can accurately specify the treatment target.Based on RGB camera and depth camera,the facial feature points of patients in the treatment process are detected in real time to calculate the position of the patient’s head,so that the TMS treatment manipulator can ensure tracking and positioning to the treatment target,thus realizing the accuracy and automation of the whole TMS treatment process.In order to solve the problem of long time-consuming and more wrong matching points,KD tree and ratiotest are used to improve.In order to solve the problem of large error in the final reconstruction of sparse point cloud,bundleadjust is used to optimize the point cloud.Finally,the 3D model of the patient’s head was obtained by surface reconstruction based on PMVs and Poisson algorithm.In order to solve the problem of tracking and locating,face detection based on LBP feature is adopted in this thesis.After the face area is detected,the CLM model is used to search and fit the feature points.Then,the nose tip is used as the coordinate origin of the detected feature points.Combined with the depth information of the depth camera and the hand eye calibration results,the face coordinate system relative to the manipulator base can be obtained Coordinate the position and pose of the system to realize tracking and positioning.The UR3 manipulator is used as the actuator,combined with the depth camera,and the upper computer software is designed,and the experimental platform is built.Compared with the traditional algorithm,the effectiveness and improvement of the proposed algorithm are verified.At the same time,it can effectively track the focus points in the treatment process,and verify that the tracking accuracy meets the standard.At the same time,in the upper computer software platform,we can observe the status of the mechanical arm and other treatment information in real time,and realize the visualization and automation of the treatment process.