3D Position Based Human Movement Computation Using Multiple Images
Jihun Park · Applied Mechanics and Materials · 2017
This paper presents a computation method of human movement using 3D point, regarding a human as a rigid body. The movement computation method uses ray vectors cast from cameras. Ray vectors cast from cameras to feature points carry values of camera external parameter. Given four or more nonplanar known points of one input image, we calculate camera's external parameters of the input image using computed values from partially overlapping, adjacent input image frames using Newton's root finding algorithm. This is achieved by computing fixed points in the environment, camera distortion values and external parameters from stationary scenes, and camera external parameters of the input frame. Using computed camera external parameters, a tracked object's rigid object movement is computed using projected intersection points between ray vectors. Our method is demonstrated using various input images. The result is used in a human tracking.