2-D kinematics of the shank and foot complex during stance using markerless segmentation and body-segment anatomical axes identification

In this work, a markerless motion capture method is proposed for the 2-D description of the movement of the shank and foot complex. The video sequences of subjects walking wearing socks or barefoot, were segmented isolating the moving regions from the background. Anatomical axes of a three-segment model of the shank and foot complex were identified using two different methods applied to the segmented images: 1) skeletonization and line detection and 2) cross-correlation of selected image regions. The axes identified were used to determine the kinematics on the sagittal plane of the two joints included in the model. Results from the segmentation of the images are independent from the presence of the socks, allowing the definition of more flexible acquisition protocols. Joint kinematics estimation with both methods appeared to be comparable to that obtained with more traditional marker-based methods. The cross-correlation method, which includes the calibration of the anatomy of the body-segments, may provide more robust solutions for the extension to a 3-D analysis of the shank and foot complex kinematics compared to the other proposed method.