Analysis of backlight images for spray measurement: how to accurately identify the liquid-gas interface in images

J. Blaisot (1)

1 CORIA - Complexe de recherche interprofessionnel en aérothermochimie

We present in this paper the key points for using backlight images in spray measurement. The main features of a spray accessible from backlight images are the interface contours of the liquid elements. It is commonly considered that using this kind of images is straightforward, notably because these images are directly understandable by a human being. However, making image-based measurement needs a rigorous approach. The first point is the setting up of the optical system. Although no special expertise is required to make images of sprays, a particular attention must be put on the imaging system and light source arrangements when relevant measurement are expected. Also, different steps must be considered to correctly estimate liquid-gas interface localization on backlight images. Fist, attention must be put on the pretreatment of the images to correct light source defects such as spatial or temporal in-homogeneity. For the segmentation step that follows, local intensity variations in the images are analyzed to determine the localization of the liquid-gas interface contour in the images. It is shown that modeling the image of liquid elements is mandatory to correctly estimate this features. Based on the results of this modeling, several local parameters are proposed. It is shown how the local shape of the interface acts on the variation of these parameters. Finally, partial overlapping of images of liquid elements is considered and solutions for discriminating potential overlapping liquid elements in an image are proposed. Introduction Image analysis is among the most promising measurement methods for achieving improved characterization of sprays. Images result from the projection of 3D objects onto 2D image plane. As such, information relative to the third direction are lost or at least hidden. However, by appropriately analyzing images, some of these information can be recovered. Spray measurements are no longer confined to the spray drop size distribution but comprise now the objective to quantify more deeply the liquid-gas interface properties. First attempts were oriented to the determination of shape parameters but the objective is now to quantify the surface area of the interface of every liquid elements in a spray, whatever their shape. To do that, an accurate determination of the interface location is needed. Backlight imaging techniques are particularly well suited for this purpose. Indeed, at least the shape of the liquid elements directly appears on the images. However, all necessary precautions are not always taken in experiments, and this is mainly due to preconceived ideas about images. First, an imaging setup may seem easier to arrange as it does not require a rigorous optical alignment to obtain the image. The problem arise when starting to use this image to make the measurement. Second, the image is directly interpretable by a human being. This easy perception gives rise to the feeling that image will be measurable as soon as image can be "understood" to the naked eye. This is far from it as a measure needs signal processing operations that may fail in case of bad images. One important drawback in using images comes from the depth-of-field (DOF) of the optical system. A spray is three dimensional by nature so images of liquid elements are focused or not, depending of the location of these elements relative to the focus plane. In addition, when spray drop size distribution is considered, the drops must be counted in a given control volume, sorting the droplets by accounting their distance from the focus plane. This must be done through the use of objective criteria determined from signal modeling, as done for all other measurement techniques. A review of the key points of the optical setup is first addressed. Important steps in image analysis operations are then presented. Finally, some issues relative to the determination of the liquid-gas interface location in images are tackled.

Domaines: Mécanique des fluides [physics.class-ph]; Traitement du signal et de l'image [eess.SP]

J. Blaisot. Analysis of backlight images for spray measurement: how to accurately identify the liquid-gas interface in images. ICLASS 2018, Jul 2018, Chicago, United States. ⟨hal-02371921⟩