@article{Greenberg_McPhee_Wong_2016, title={Quasi-Monte and Data-Driven Monte Carlo Methods for Efficient Human Joint Model Fitting}, volume={2}, url={https://openjournals.uwaterloo.ca/index.php/vsl/article/view/124}, DOI={10.15353/vsnl.v2i1.124}, abstractNote={<p>Fitting a kinematic model of the human body to an image without<br />the use of markers is a method of pose estimation that is useful<br />for tracking and posture evaluation. This model-fitting is challenging<br />due to the variation in human physique and the large number<br />of possible poses. One type of modeling is to represent the human<br />body as a set of rigid body volumes. These volumes can be<br />registered to a target point cloud acquired from a depth camera<br />using the Iterative Closest Point (ICP) algorithm. The speed of ICP<br />registration is inversely proportional to the number of points in the<br />model and the target point clouds, and using the entire target point<br />cloud in this registration is too slow for real-time applications. This<br />work proposes the use of data-driven Monte Carlo methods to select<br />a subset of points from the target point cloud that maintains or<br />improves the accuracy of the point cloud registration for joint localization<br />in real time. For this application, we investigate curvature of<br />the depth image as the driving variable to guide the sampling, and<br />compare it with benchmark random sampling techniques.</p>}, number={1}, journal={Journal of Computational Vision and Imaging Systems}, author={Greenberg, Sara and McPhee, John and Wong, Alexander}, year={2016}, month={Oct.} }