Content-based surface material retrieval
Year: 2017
Type: conference-paper
Abstract: We present a content-based surface material retrieval (CBSMR) system for tool-mediated freehand surface exploration that relies on features motivated by the main psychophysical dimensions of tactile surface texture perception. The proposed approach does not require explicit scan force and scan velocity measurements. The perceptual features used in our CBSMR engine cover the tactile dimensions of friction, hardness, macroscopic roughness, microscopic roughness and warmth. We examine 108 surface materials recorded by different users and present the results of a free-sorting grouping experiment with 30 subjects which we conducted to determine the perceptual similarity of the surface materials in our database, providing a ground truth data set for perceived tactile similarity. The outcome of this experiment is used to demonstrate that the proposed CBSMR engine is able to determine the perceptually most similar surface materials for a test query. The proposed set of 8 features leads to a classification precision of 86% and a similarity precision-at-one of 30% when combined with a Euclidean Distance-based classifier.
Authors: Matti Strese, Yannik Boeck, Eckehard Steinbach
Institution: Technical University of Munich
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2740112960
