Presentation
Biomechanical Analysis Using a Computer Vision-Based Motion Capture
DescriptionRepetitive muscle use significantly impacts human bodies, often as a long-term aftereffect of poor
posture, heavy load handling, and other strenuous practices. Biomechanical software has emerged as a vital tool in understanding the performance of human muscles under certain load conditions, being used to develop a safe threshold of the human body’s range of muscle use. This research has implications for prevention of work-related injury as well as long-term care practice. Paired with motion capture systems to record the human body’s motion when interacting with their immediate environment, biomechanical software enables the virtual visualization of the captured actions. A user input of the various forces acting on the human limbs during interaction with objects in their surroundings or while performing strenuous tasks allows real-time simulation of the loads experienced by individual muscles, thus determining if the action is feasible without causing muscle trauma or injury.
This study introduces two camera-based motion capture methods and their integration with
biomechanical software. The first method requires the physical markers being placed on the
subject’s body that are captured by a set of cameras, such as Vicon system. The second method
applies an image processing method, such as Media Pipe Pose Detection Kit, without the physical
markers. In the second method, only a single camera is used to detect the human body's posture
and automatically generate the virtual markers on the body to detect the limbs and joints. The
resultant data presents the location of each marker in space for the duration of the motion.
While there are several biomechanics software available, such as AnyBody, OpenSIM, the one
employed in our research is called Biomechanics of Bodies, or BoB, which offers a full-body
musculoskeletal model, as others do not. While BoB can work with Vicon, it adopts different
marker placements from MediaPipe. Therefore, it is essential to develop a method to convert the
MediaPipe results into a data format recognizable by BoB. The authors’ intention to create an
open-source solution has resulted in the development of a data conversion pipeline from the marker position results obtained from MediaPipe to be input into BoB for the musculoskeletal model visualization and biomechanical analysis. An open-source Python kit, called Kinetics Toolkit, is employed as an intermediary data converter, along with a series of 3D transformation matrices based on Helen-Hayes system to display the appropriate pose for the simulation. Compared to Vicon system, the MediaPipe system, once developed, would be more cost effective and applicable for public health care uses.
posture, heavy load handling, and other strenuous practices. Biomechanical software has emerged as a vital tool in understanding the performance of human muscles under certain load conditions, being used to develop a safe threshold of the human body’s range of muscle use. This research has implications for prevention of work-related injury as well as long-term care practice. Paired with motion capture systems to record the human body’s motion when interacting with their immediate environment, biomechanical software enables the virtual visualization of the captured actions. A user input of the various forces acting on the human limbs during interaction with objects in their surroundings or while performing strenuous tasks allows real-time simulation of the loads experienced by individual muscles, thus determining if the action is feasible without causing muscle trauma or injury.
This study introduces two camera-based motion capture methods and their integration with
biomechanical software. The first method requires the physical markers being placed on the
subject’s body that are captured by a set of cameras, such as Vicon system. The second method
applies an image processing method, such as Media Pipe Pose Detection Kit, without the physical
markers. In the second method, only a single camera is used to detect the human body's posture
and automatically generate the virtual markers on the body to detect the limbs and joints. The
resultant data presents the location of each marker in space for the duration of the motion.
While there are several biomechanics software available, such as AnyBody, OpenSIM, the one
employed in our research is called Biomechanics of Bodies, or BoB, which offers a full-body
musculoskeletal model, as others do not. While BoB can work with Vicon, it adopts different
marker placements from MediaPipe. Therefore, it is essential to develop a method to convert the
MediaPipe results into a data format recognizable by BoB. The authors’ intention to create an
open-source solution has resulted in the development of a data conversion pipeline from the marker position results obtained from MediaPipe to be input into BoB for the musculoskeletal model visualization and biomechanical analysis. An open-source Python kit, called Kinetics Toolkit, is employed as an intermediary data converter, along with a series of 3D transformation matrices based on Helen-Hayes system to display the appropriate pose for the simulation. Compared to Vicon system, the MediaPipe system, once developed, would be more cost effective and applicable for public health care uses.
Event Type
Oral Presentations
TimeTuesday, April 13:50pm - 4:10pm EDT
LocationPier 9
Simulation and Education (SE)