THE APPLICATION OF MACHINE VISION TO THE AUTOMATED INSPECTION OF KNITTED FABRICS

The development of modern Charge Coupled Device (CCD) image sensing devices and the falling costs and rising power of modern computers mean that the automation of mundane and repetitive visual inspection tasks is becoming feasible at a low cost. In particular, low cost automated fabric inspection has become feasible due to the development of the linescan camera, and this device is used as the basis for this project. Previous commercial fabric inspection systems have been very expensive, and an overriding objective of this work is to produce a very low cost solution. This has been achieved by using commercially available components wherever possible, with the exception of the camera interface/processor board which has been specially fabricated. A processing methodology has been developed that is essentially a strategy for data reduction. The first stage is a coarse operation that quickly removes in excess of 99% of the data. This is possible because of the assumption that the vast majority of the fabric is defect free and can be safely ignored. If this were not so, the system would become unstable and inoperable. Further processing then takes place on the smaller quantities of data, and the algorithms can therefore be more complex and computationally expensive. Ultimately, the data is transformed into a pass/fail indication along with a defect classification.