Light stream method for improving robustness of particle image velocimetry based on physics

Abstract

The invention provides a robust light stream calculation method based on the physics so as to improve the light stream calculation robustness in allusion to a problem that edge diffusion of a visual flow image and influences of noise and abnormal points enable the light stream calculation robustness to be poor. The algorithm introduces an anisotropic filter into a physics based light stream methodso as to enhance the edge light stream robustness, increases a penalty factor so as to reduce the influences imposed on light stream calculation by noise and abnormal points, then minimizes a light stream energy function based on a variational method so as to solve an Euler-Lagrange equation, and finally solves a velocity field according to an iterative method. Simulation results show that compared with the traditional Lucas-Kanade, Horn-Schunck, pyramid Lucas-Kanade and physics based light stream calculation methods, the algorithm provided by the invention can significantly reduce light stream diffusion in edge and corner regions and improve the robustness in allusion to the noise and abnormal points, so that the velocity field with good robustness is acquired.

Description

technical field [0001] The invention belongs to the field of unmanned aerial vehicle video detection, and in particular relates to a physics-based optical flow method for improving the robustness of particle image velocity measurement. Background technique [0002] Fluid velocity field measurement is of great significance for understanding complex fluids in fluid mechanics and aerodynamics research. Fluid motion is a typical non-rigid motion, and its calculation needs to be based on image processing technology. The size, direction and distribution of the local fluid motion vector can be obtained by analyzing the moving image sequence, and then physical characteristics such as viscosity and eddy current field distribution can be obtained. Usually, since the moving object is transparent or difficult to observe through optical equipment, it is necessary to place visible particles into the object to obtain the fluid motion characteristics indirectly by estimating the particle m...

AI Technical Summary

Problems solved by technology

[0004] Ma Pengfei et al. proposed a PIV estimation method based on Lucsa-kanade (LK) local optical flow. This algorithm has low computational complexity and good robustness. However, the boundary of the optical flow field is blurred, and the optical flow estimation of edge pixels is poor. The flow field is relatively sparse

In response to this problem, Sun Lizhi et al. proposed a multi-scale optical flow algorithm based on pyramid LK to calculate PIV. The proposed algorithm can solve the problem of optical flow calculation when the speed of the moving target is too high and the continuity of adjacent frames is not strong. The obtained optical flow accuracy Higher than the LK algorithm, but the obtained optical flow field is still relatively sparse

For this reason, Huang Zhan et al. proposed a PIV calculation method based on Horn-Schunck (HS). This algorithm can obtain a dense optical flow field, but the boundary is easy to blur, the calculation complexity is high, and the robustness is poor.

In order to further improve the accuracy of optical flow estimation, Yu Xuemin et al. based on the combination of global and local methods to obtain higher optical flow estimation accuracy, but the optical flow obtained by this method is less robust in the edge area

Tianshu L. and other physics-based optical flow methods obtain high-resolution velocity fields from visualized images. However, due to the diffusion of optical flow in the edge area and the influence of noise and outliers, the robustness of optical flow is poor

Images