Defect detection for patterned textile fabrics with deformations

Deformations of patterns during textile fabric manufacturing are very common in practice. However, acceptable deformations may easily be treated as defects which will cause false alarms by most defect detection schemes. In this paper, a novel defect detection scheme is proposed for patterned textile fabrics which exhibit deformations of patterns. The defect detection problem is initially formulated as a nonlinear functional optimization problem. Upon parametrization of the functional, the problem can be transformed into a general optimization problem with the objective of selecting the parameters (i.e., translations along the x, y directions and the rotation angle) and the deformed functions such that the deviation between the defective image and the feature image is minimized. Several examples are used to illustrate the effectiveness of the proposed method. Experimental results obtained clearly show that false alarms can be avoided and any potential defects can be isolated and detected without ambiguity.