Automated wear label assessment in carpets by using local binary pattern statistics on depth and intensity images

Abstract

Carpet customers want a product of which the appearance lasts for years. Therefore, carpet manufacturers certify their products with labels that represent the expected change in appearance after the first year of installation. No automated system exists yet for objectively assigning these ranks. In this approach, we present an automated method for assessing carpet wear based on image analysis. For this, depth and intensity information are captured from eight types of carpet samples. The results show that the method correctly assigns wear labels from 1 to 5 in steps of 1 for six of the eight carpet types.