| Use robots to assist or replace doctors to carry out reduction surgery for patients with ankle injuries can greatly reduce the workload of doctors.and the high stability and motion accuracy of robots are also conducive to improving the effect of orthopedic surgery.Before performing orthopedic surgery,the foot and aukle orthopedic robot needs to obtain the position and posture information of the patient’s feel,laying the foundation for the robot to further grasp and perform surgical operations,Because of machine vision has the advantages of simple deployment and wide perception range,It is an important means for foot and ankle orthopedic robots to obtain seene information,and plays an important role in identifying human foot largets,calculating the accurate posture of the target,guiding the robot to move to the specilied position,and completing the preoperative guidance of the robot.In this paper,we carry out the research on visual detection,positioning and preoperative guidance for the application of foot and ankle orthopedic robots.and the main research contents are as follws:(1)Explore the image segmentation method of human feet.used the obtained image infurmation of the surgical scene,and the traditional digital image mcthods such as edge detection and area search were used to realize the segmentation of foot features.Aiming at the shortcomings of iraditional methods,combined with self-built datasets,the application scbeme of Mask RCNN model based on deep convolutional network in foot image segmentation is further proposed and implemented,and the proposed method is compared with the traditional method,and the results show that the machine learning method has obvious advantages in resisting the background environmental impact.(2)Establish a point cloud segmentation and positioning method combinig images.Through the joint calibration of RGB-D camera system,the camera RGB image and depth image registration is completed.and the method of extracting the target poiny cloud by using the image segmentation result is realized to solve the problem of large calculation of traditional point cloud segmentation.Combined with point cloud processing and PCA method,a pose estimation method for target point cloud data is proposed.and further combined with directed eneitelement box and human foot sign statistics,the spatial pose results of foot guidance target points under the camera coordinate system optimized,and the optimized guidance target pose meets the requirements of robot guidance tasks.(3)ealize robot motion guidance based on visual positioning and safety detection.The coordinate system converson relationship between the camera system and the end of the robot is established by hand-eye calibration,and the forward and reverse kinematics model of the robot is establishccl by combining the D-H parameters to solve the current state of the robot.Realize motion guidance hased on the position and attitude change of the motion target and the rotation of each joint of the robot.The safety feedback method of the robot preoperative guidance scene was established,the robot motion detection method and the patient’s foot abnormal movement detection method were realized to guide the robot movement,and the robot motion guidance method combined with safety detection was established.(4)Carry out prototype experimental platform construction and technical verification.According to the task requirements,the experimental platform of osteopathic assisted robot was built.the upper computer control software was developed based on the PyQt5 frame work,and a prototype system of foot and ankle orthopedic robot with complete functions was built in combination with hardware equipment,and experunental verification was carried out for key technologies such as image segmentation model training,point cloud segmentation and positioning.and robot guidance.Relevant experimental results show that the proposed method has good applicability.In this paper,a complete human foot image-point cloud fusion detection and positioning framework is established for the cucurrent foot detection and positioning of foot and ankle orthopedic robots,which realizes the functions of preoperative visual positioning and motion guidance of the robot,and verities the application of key technologies based on the prototype experimental platform built,and the results show that the proposed method has high foot target detection and positioning ability,which provides theoretical and practical guidance for the further application of preoperative guidance of orthopedie robots.Robot technology has been widely applied in the aerospace,medical and other fields.the perception of external environment and target discrimination are important indicators to evaluate the intelligence of robot.However,traditional robot external perception methods such as infrarer,laser.ultrasonic and on often have limitations in perceiving the environment range and obtaining insufficient information.Most of them are unable to satisfy the demands of complicated situations like robotic assistance.with the rapid advances in the technology of computer recently,machine vision has been applied to a wide range of applications,such as scene awareness and robot guidance.This paper takes ankle-assisted orthopedic robot as the research object,and establishes a human foot target detection,localization and robot guidance method based on image-point cloud fusion for the preoperative patient foot target detection and guidance problem of orthopedic assistant robot.In this paper,we focus on the following issues:the recognition of human foot target based on RGB images,the segmentation and space posture computation of the target point cloud,the Hand-eye calibration and the 6-DOF robot orientation,etc.The main research contents are as follows:... |
PDF Full Text Request