Algorithm for high-accuracy particle image position estimation in PIV applications

ABSTRACT We proposed a method for the 3D position estimation in particle image velocimetry. The method uses the pattern–matching between theoretical and experimental images by exploiting the scattered energy field and uses genetic algorithm. The simulations and experimental verification of this problem are discussed. Keywords: Particle image velocimetry, genetic algorithm. 1.- INTRODUCTION In particle image velocimetry is important to obtain information about the three-component positions. There are several works in order to provide instantaneous three-dimensional position information, such as holographic, stereoscopic and light sheets methods[1-2]. For practical applications, restricted optical access often eliminates stereoscopic approaches. Robustness, experimental facilities, and the need for real-time results make holography an unattractive option. In other hand the scanning light-sheets are difficult to operate with restricted optical access in industrial applications. Velocimetry particle images show a scattering field that is dependent on their relative 3D position when illuminated in a volume, such as when holographically recorded or imaged using Tunnelling Velocimetry [3]. The Tunnelling Velocimetry technique involves in-line illumination of a volume of interest, achieved through a single instrument using a single optical access point, thus obtaining seeding particle scattering images produced within said volume. In this work the experimental image was obtained with Tunnelling Velocimetry technique and was interpreted to obtain the 3D information of particle position. So, the diffraction field can be deduced from single CCD camera[4] position for a range of x,y,z particle position , opening the way for using a single camera in practical experiments. In other hand, the theoretical image was generated by treatment Generalized Lorentz-Mie theory (GLMT)[5]. The particle positioning method using GLMT to create theoretical seeding particle images, and embedding the genetic algorithm search concept, was devised and is described in the present work.