REAL-TIME QUALITY CONTROL OF FABRIC BASED ON MULTIVARIATE CONTROL CHARTS

This contribution deals with a procedure of monitoring fabric production process in terms of defect detection. Detection algorithm is based on spectral approach with 2D Fourier transform, which has shown to be suitable method of describing periodic and almost periodic structure of fabrics. Multivariate control charts are used as a tool for monitoring several quality variables. In order to make an off-line and on-line quality monitoring of fabrics possible, we built up a prototype machine, which consists of aluminium frame, DC motor, LED lights and a line scan camera. Fabric width is 50cm and maximum winding speed of about 16m/min can be achieved. The core of detection algorithm utilizes relatively time consuming Fourier analysis, so we split an acquired image into independent parts and let the work be done concurrently. Writing native, multi-threaded and scalable code, that takes advantage of modern multi-core machines, is one of our main objectives. Based on Visual C++ language and additional libraries, we are able to to check 1m 2 of fabric in about 1.5 second on Intel Core2 Quad CPU .W e use OpenCV library together with Intel Threading Building Blocks templates for the most part.