Hybrid Approach using correlation and morphological approaches for GFDD of plain weave fabric

This paper presents the Hybrid Approach (HA) using correlation followed by Morphological Approach (MA) to detect the micro natured defects in fine grey plain weave fabrics. The fundamental method of suspected object detection in an image is achieved using Correlation Approach (CA) to determine correlation of defect template with defect image to detect Region of Interest (ROI). The FFT based correlation technique assisted by appropriate threshold was followed by the MA using suitable Structuring Element (SE) to filter and retain only the defect region and to facilitate the reduction of False Alarm Rate (FAR). Three methods viz., MA, CA and HA were tested and compared for Grey Fabric Defect Detection(GFDD) of plain weave defects of varying sizes. Performance comparison of the three algorithms was carried out by adopting simple binary based Defect Search Algorithm (DSA) as a last step in the experimentation to detect defects. The progressive improvement observed in terms of % Overall Detection Accuracy (ODA) for these three methods was from 22 to 79 for benchmark defect samples and from 25 to 93 for normal samples. Experimentation results on HA showed 3.5 fold increase in ODA for defect samples and about 3.7 fold increase in ODA for normal samples. Though HA method could detect different defects, the drawbacks of this are the selection of the appropriate defect template and selection of SE which is tedious. The details of the experimentation and the results thereof are presented in this paper.