Nowadays, rapid development in the technology of optical and optoelectronic detectors leads the usage of the computer vision based feature extraction techniques which are examined in a wide range from forensic investigations to automatic quality control of industrial products. Automatic identification of the production fault and production parameters of textile materials is one of the investigation fields, in which the computer vision methods become crucial in terms of the prevention of loss of raw material and labor time. In this study, in order to determine the twist level of Chenille yarn, which is extensively used in woven and knitted fabrics of our daily life, segmented image in which the component yarns are segmented in different scales of grey-level values is obtained by using sequential image processing algorithms. Besides the statistical examination of the axial grey-level signal extracted from segmented image, texture correlation curve determined by using Haralick's GLCM (grey-level co-occurence matrix) features also ensures the successfully identification of twist level of the yarn. Then, Hough Transform (HT) has been used to predict the reflection and barré faults which are encountered in the fabrics to be produced from the yarns whose pile direction has not been adjusted after winding process in accordance with the original production direction.