A detection of tears in laces using image processing

High quality laces must contain no defects. However, sometimes tears appear after or during production process. These tears are usually detected by humans. This paper proposes image processing technique that can automatically detect possible tears in order to help humans detecting tears in lace faster. The proposed method is based on the idea of segmenting a lace image into parts based on the size of holes inside the lace texture. Assumming that tears occurs only in the parts with holes, the part with no holes are elimiated first by binary erosion method. Then, the result is used to cut out only parts with holes from the oringinal image. Then, the image of only parts with holes are binarized and logically nor with the parts with no holes so that the holes become white closed areas. Then, after applying the labelling technique, parts with the same size of holes are extracted by applying the binary open morphology from smaller structure size to larger one. Finally, tears in the parts with holes of the same size are detected by eliminated the holes with correct size assuming that tears causes the holes to be larger. The proposed method was tested with images of lace samples with errors of 0.81 %.