Real-Time Surface Flaw Detection for the Leather and Textile Industries Using Machine Vision Techniques.

Requirements for the visual inspection of flat surfaces exist in both the leather and textile industries. However to date (1995) there has not been wide-scale use of automatic visual inspection within these industries, primarily for two reasons: the surfaces which typically arise within the industries have diverse visual characteristics the companies operating within the industries cannot afford the automatic visual inspection technology which is currently available for other industries. Recently there have been rapid developments in three key areas of technology all of which are relevant to the cost-effective implementation of automatic visual inspection: cameras programmable processors programmable logic. The work described in this thesis has sought to determine how these developments can be exploited to meet the requirements of the selected industries. Two techniques have been developed for the detection of defects in textiles, one based on suppression of the textile's weave the other on texture analysis. The first of these is suited to applications in which it is necessary to detect defects which have a size comparable to that of the spacing of the threads in the weave, while the second suitable for the detection of large and extended defects. In addition, computationally efficient processing techniques have been formulated for the enhancement and detection of defects on leather. The implementation of these techniques has been investigated through the development of a general purpose surface inspection testbed employing a combination of high definition linescan cameras, high performance digital signal processors (Texas Instrument's TMS320C40s) and programmable logic. It has been concluded that it is possible to meet the requirements which the selected industries have for automatic visual inspection in a cost-effective manner. It has also been concluded that while the automatic visual inspection of surfaces is apparently a generic task with application in many manufacturing processes, for an implementation to give an acceptable level of performance, it is necessary to utilise image acquisition and digital image processing techniques which are well matched to the visual characteristics of the surface to be inspected