Moving Objects, Mesh-based Motion Estimation Methods

In the field of aviation surveying, it is a very important research section of the untouchable surveying on the moving parameter of a moving object. Because in the performance measurement of the aircraft, thruster, weapon system, wind tunnel field stream etc., it is common to meet the circumstance that it's uneasy to use the touchable sensor on the measurement. With the fast development of digital video processing, computer vision and onboard computer technology, it makes possible the implementation of untouchable surveying using digital video processing analysis in combination with deformation measure model.This paper makes some research on the problems as the motion parameter, follow-up tracing and estimation of a moving object on video frequency. It also puts forward a kind of motion estimation method with the aim to video sequence image, this method is on the basis of the net model. The main content and the achieved research outcome of this article are just as follows:1). A sort of generating algorithm on the basis of the mesh model. First, it gets the border points of a video moving target by the polygon approximate, and uses SUNS AN testing operator check the angle points and uses difference of a function two frame informations to eliminate some points. Then take these points as the mesh points of Delaunay triangular mesh model to make up the net.2). Second, calculate the scattered optic flow file so as to determine the optic flow of the mesh node. In the motion estimation of the mesh model, the estimation way of the mesh node is a special important step, and the estimation must has a higher Robust character, otherwise, it will cause a bigger change on the mesh capture structure of the succeed frames, as a result, some serious joint effect such as the compensation on the succeed movement and the reconstruction of the predicated picture will happen. We adopt the method that is probabilistic, using Bayesian estimation, in particular MAP estimation, which would better and improve the estimation precision of the moving vector by a large margin.3). Last, calculate the dense optic flow field so as to make sure the stream of all the pixels. In the demonstration of the mesh model, mesh's moving are generally described by the net apex. For the similar characteristics between the apex and the internal points, the moving vector of the internal points is usually got by the interpolation of the net node.