Soft sensor for the characterization of fibrous materials through multivariate and multiresolution image analysis

In fiber technology, the characterization of both morphology and physical features of fiber fabrics is of paramount importance. Usually, the key characteristics of fibrous materials are determined by time-consuming and possibly destructive laboratory analysis. In this paper, a multivariate and multiresolution software sensor is presented that exploits the information embedded in images of fibrous materials to measure the morphology of the product and also to estimate its morphological and physical characteristics. The soft sensor includes two artificial vision systems performing: 1) the direct measurement of the fiber diameter distribution and 2) the estimation of fiber diameter distribution and permeability from texture analysis. The soft sensor is tested in the case of scanning electron microscope images of polymer nanofiber membranes, and demonstrates to be accurate and reliable.