Artiﬁcial Vision System for the Automatic Measurement of Interﬁber Pore Characteristics and Fiber Diameter Distribution in Nanoﬁber Assemblies Emanuele Tomba, † Pierantonio Facco, † Martina Roso, ‡ Michele Modesti, ‡ Fabrizio Bezzo, † and Massimiliano Barolo* ,† Computer-Aided Process Engineering Laboratory (CAPE-Lab), Dipartimento di Principi e Impianti di Ingegneria Chimica, and Dipartimento di Processi Chimici dell’Ingegneria, UniVersita ` di PadoVa, Via Marzolo 9, 35131 PadoVa PD, Italy Nanoﬁber structures are used in several technologies such as membranes, reinforced materials, textiles, catalysts, sensors, and biomedical materials. For all these applications, it is important to know the morphology of the assemblies, in particular their pore and ﬁber dimension distributions. However, the current methods used to measure pore sizes are all experimental and indirect; furthermore, the ﬁber diameter distribution is usually determined manually using a digital image of the nanoﬁber web. In this paper an artiﬁcial vision system is proposed to characterize the nanoﬁber web by automatically measuring several properties related to the interﬁber pore distribution and to the nanoﬁber diameter distribution. The artiﬁcial vision system is characterized by a two-section structure: an image processing section and a property measurement section. The image processing section is centered on a multivariate image analysis procedure for the extraction of morphological features from the image. The property measurement section comprises an algorithm for interﬁber pore area and pore morphology evaluation and one for ﬁber diameter distribution measurement that also accounts for the effect of perspective on the lower-level ﬁber diameters. Because the proposed artiﬁcial vision system is completely automatic, measurements can be taken without the need of any experimental setup and with no human intervention. Therefore, besides being fast and accurate, measurements do not suffer from repeatability issues. The ability of the proposed automatic system in characterizing the morphology of a thin nonwoven nanoﬁber fabric is demonstrated by application to polymer nanoﬁber membranes obtained by electrospinning. 1. Introduction Nanoﬁber membranes combine the advantage of high total porosity with low dimensions of pores. These peculiar features have been exploited in several applications, such as ﬁltration and separation, 1 catalysis, 2,3 and biomedical materials process- ing. 4 In addition to this, the nanometric dimension of ﬁbers leads to important advantages, such as increase of the area-to-volume ratio, greater surface functionality, and better mechanical performance. In particular, the dimension of nanoﬁber diameters is crucial in those applications where high speciﬁc area is required, such as in tissue engineering and drug delivery systems, 5,6 for applications in the reinforcement of composite materials, 7 in the production of protective clothing, 8 or in applications as high-sensitivity nanosensors. 9 The suitability of a nanoﬁber membrane for the application it has been manufactured for is therefore strongly related to the interﬁber pore and the ﬁber diameter morphologies. However, as in most nanomaterials, these morphologies are difﬁcult to characterize in a systematic way. Measurement instrumentation may not be available or may require processing experimentally a sample of the material being characterized (possibly with destructive tests), and almost invariably calls for human intervention, which may lead to loss of repeatability and possibly large measurement times. Porosity is measured experimentally by mercury intrusion porosimetry, liquid extrusion porosimetry, or capillary ﬂow porosimetry. 10 Mercury intrusion porosimetry requires pushing mercury through pores in the material, but the high pressure required for this operation may distort the nanoﬁbers, thus compromising the mat morphology. Liquid extrusion porosim- etry is suitable for testing nanoﬁber membranes, but it measures the total pore volume, rather than the pore throat diameter, which instead is the property one is most interested in. Capillary ﬂow porosimetry does provide a simple and nondestructive technique allowing for a rapid and accurate measurement of pore size and distribution. 11 However, being itself an experimental method, operators and time (hence money) are needed; moreover, it is not always accurate and may downgrade the sample. Contrary to pore size distribution, experimental methods for nanoﬁber diameter measurement are not available at present. Instead, the ﬁber diameter size is determined by visual inspection of scanning electron microscope (SEM) images of the nanoﬁber mat. Typically, the diameter dimension is measured by manually counting the pixels between ﬁber boundaries after a ﬁber has been visually isolated by the operator on the image scene. 12 It is therefore clear that this time-consuming procedure is subject to the operator’s judgment, is quite prone to measurement errors, and provides hardly reproducible measurements. The availability of digital microscopic images of complex nanoﬁber webs provides an attractive basis for the development of an automatic system for nanoﬁber characterization, where no human intervention is needed. Such an artiﬁcial vision system would be nondestructive, i.e., it could capture information from the inspected product without interfering with it, thus avoiding the typical issues of human visual inspection systems. Further- more, being automatized, it would lead to perfectly reproducible results. As far as membrane porosity characterization is concerned, She and co-workers 13 applied an image analysis technique to automatically measure the effective pore diameters and their * To whom correspondence should be addressed. E-mail: max.barolo@ unipd.it. † CAPE-Lab, Dipartimento di Principi e Impianti di Ingegneria Chimica. ‡ Dipartimento di Processi Chimici dell’Ingegneria. Ind. Eng. Chem. Res. 2010, 49, 2957–2968 2957 10.1021/ie901179m  2010 American Chemical Society Published on Web 02/08/2010