Principal components analysis for PIV applications
by
Antonio Cenedese
DITS - Università di Roma "La Sapienza"
Coauthors: A. Pocecco, G. Querzoli

During last decade the experimental flow investigation based on image analysis (PIV and PTV) received a big development due to the improvement of hardware support and acquisition systems (digital cameras and image acquisition cards). PIV technique is based on the detection of the displacement of seeding particles which can represent, for their characteristics, the flow [1]. In PIV the displacement is calculated by analysing images acquired at known times, the images are subdivided in sub-domains and the mean particles displacement is calculated by means of auto or cross-correlation. By using cross-correlation it is possible to increase the dynamical range of measurement but it requires the acquisition of two distinct images. Several techniques were suggested in order to obtain two single-exposed images: - Cross-correlation cameras, which can acquire two digital images, commonly with 1000x1000 pixels. - Digital cameras equipped with narrow-band optical filters and using lasers emitting at different wavelength [2]. - A photo camera, with a laser system emitting two different colour laser beam, acquires a multiexposed colour image. A successive separation into two images utilising the colour information is required in order to apply cross-correlation technique. The present work proposes a method, based on the detection of the principal components of the image, to obtain two single-exposed images from a colour digital PIV image.

Date received: January 25, 1999