Leveraging Two Kinect Sensors For Accurate Full-Body Motion Capture

It is human motion capture which means restoring human's pose parameters from a single perspective or multiple synchronous or asynchronous angle data or a sequence.It's fundamental to keep the accuracy and robustness of motion capture.For a long time,many scholars and engineers are working on advanced motion capture algorithm,which means they want to reconstruct human motion in 3D space not only real-time but also automatically and accurately.Accurate motion capture is widely used in sports analysis,virtual reality and many other fields,which enriches people's entertainment and daily life.Currently,great progress of motion capture has been made in the aca-demic field.However,huge challenges still stay in local optimization,which encour-ages many scholars and engineers work for motion capture algorithm more actively.In this paper,we propose a complete system to measure the pose parameters of human body accurately.At first,we also focus on a sematic shape reconstruction al-gorithm.Sematic parameters and corresponding shape variations are mapped with a linear function,and we further use the captured 3D clouds to be the depth constraint to reconstruct accurate human model,which is beneficial for improving accuracy of mo-tion capture.For motion capture,it's different from previous monocular depth camera systems that we leverage two Kinect sensors to acquire more information about human movements,and the captured pose of previous frame are used to be the template of cur-rent frame and we finally transfer the pose estimation problem to be an MAP problem.Usage of two Kinect sensors ensures that we can still get an accurate estimation even when significant occlusion occurs.Because human motion is temporally constant,we adopt a learning analysis to mine the temporal information across the posture variation-s.Using this information,we estimate human pose parameters accurately,regardless of rapid movement.Our experimental results show that our system can perform an accurate pose estimation of human body with the constraint of information.