Method for Regularized Estimation of the Optical Flow Velocity Field in Computer Vision Systems

For computer vision systems that use monocular methods for determining the motion of a video camera based on algorithms for calculating optical flow parameters are proposed method velocity field estimates that are resistant to spatiotemporal interference. The current estimate of the optical flow parameters is formed by minimization of the integrand augmented energy functionality regularizing components: according to Tikhonov and taking into account the continuity of the velocity field in time. In comparison with the known procedure for estimating the velocity field by minimizing full energy functionality using optical flow values for the entire scanning area and requiring a significant amount of calculations, the implementation of the proposed method uses optical flow measurements at only one point in the scene. The simple analytical form of the obtained solution is easily implemented in on-board computers. The conducted numerical experiment illustrates the possibility of practical application of the proposed approach in computer vision systems ensuring the functioning of the systems spatial orientation and positioning of autonomous devices general purpose robots, Earth surface sensing complexes electrodynamic, mechanical, and other systems.