Image processing of vaporizing GDI sprays: a new curvature-based approach

Abstract

This article introduces an innovative method for the segmentation of Mie-scattering and schlieren images of GDI sprays. The contours of the liquid phase are obtained by segmenting the scattering images of the spray by means of optimal filtering of the image, relying on variational methods, and an original thresholding procedure based on an iterative application of Otsu’s method. The segmentation of schlieren images, to get the contours of the spray vapour phase, is obtained by exploiting the surface curvature of the image to strongly enhance the intensity texture due to the vapour density gradients. This approach allows one to unambiguously discern the whole vapour phase of the spray from the background. Additional information about the spray liquid phase can be obtained by thresholding filtered schlieren images. The potential of this method has been substantiated in the segmentation of schlieren and scattering images of a GDI spray of isooctane. The fuel, heated to 363 K, was injected into nitrogen at a density of 1.12 and 3.5 kg m−3 with temperatures of 333 K and 573 K.

Characterization of the liquid phase of vaporizing GDI sprays from Schlieren imaging

“…The experimental setup will be briefly described below, refer ring to previous papers [24,26] for further details. The char acterization of isooctane sprays was carried out in a constant volume vessel under inert environment, using the 8holes injector manufactured by Delphi according the specifications of the engine combustion network (ECN) [12,27].…”

“…Several effective methods of seg mentation have been developed [13,[28][29][30][31][32][33][34][35][36][37][38][39][40] and new ones are proposed periodically. We have recently developed an original image processing method, which has proven to be very effec tive in segmenting schlieren and scattering images of vapor izing GDI sprays [24]. As previously described, the method essentially involves three basic steps: an effective filtering of spray images through their regularization; an iterative thresh olding procedure based on the Otsu's method; the magnifica tion of the schlieren texture.…”

“…The present work aims to demonstrate how it is actually possible to distinguish the liquid phase of a vaporizing GDI spray from its schlieren image, thus allowing to characterize both spray phases in a truly simultaneous way. An optimal image processing method has been applied [24], which is essentially based on three main steps: the 'optimal' filtering of the spray images by means of their regularization through variational methods; a thresholding procedure based on the iterative application of the Otsu's method [25], the magnifi cation of the schlieren texture by means of the principal cur vatures of the spray image surface. The method was recently tested in the segmentation of the vapour and liquid phases of a vaporizing GDI spray from schlieren and Miescattering images respectively [24].…”

“…An optimal image processing method has been applied [24], which is essentially based on three main steps: the 'optimal' filtering of the spray images by means of their regularization through variational methods; a thresholding procedure based on the iterative application of the Otsu's method [25], the magnifi cation of the schlieren texture by means of the principal cur vatures of the spray image surface. The method was recently tested in the segmentation of the vapour and liquid phases of a vaporizing GDI spray from schlieren and Miescattering images respectively [24]. In the present work the method has been further refined and applied to characterize both the vapour and liquid phase from schlieren images of a GDI spray under different experimental conditions.…”

Quantifying extinction imaging of fuel sprays considering scattering errors

“…Lately there are also efforts to distinguish the liquid phase in a Schlieren setup by a filtering of the images, allowing the characterization of the liquid and vapor phase simultaneous from the same point of view. 10,14,15 All these techniques allow a good understanding of macroscopic spray propagation behavior; however, they are unable to detect quantitative information. Without this quantitative information, we cannot draw any conclusion regarding the mixture formation.…”