Abstract
We propose a new algorithm to simultaneously recover the motion and 3D shape of moving objects from feature points of the moving objects in an orthographic projection. Given the assumption that the moving object is a rigid body, the newly proposed algorithm uses times series correspondence of feature point coordinates in the orthographic projections and is based on the correspondence of 4 feature points at 3 points in time where the existence of a solution has been demonstrated. The merit of this algorithm is that it is a linear algorithm for recovery that uses only observed data, has the condition of unique recovery solution in principle, and can be applied in a unified manner in the case of a unique solution. In this method, after the possibility of recovery is determined by calculating the rank of the matrix, the recovery solution can quickly be found in a unified manner through a linear computation including matrix inversion computations. This theory solves the usual problems in the linear recovery theory in the recovery of moving objects from the point correspondence in orthographic projections and is segmented in theory. We verified the validity of this algorithm using simulation data generated by a computer, and after obtaining corresponding points, verified that realβtime processing with an average recovery time of 4.6 ms is possible even for a personal computer CPU processing. In addition, the effectiveness of this algorithm was confirmed by using the real image of a moving object.