A DSP Implementation of an AOM and its Application to Defects Detection in Textile Material.

This paper explains a method of defects detection in textile material using a DSP. This Supervised Learning method will allow the detection of defects in anyone of the phases of production. An algorithm of pattern classification based on minimum distance is used to carry out this method. Scalar distance in an Associative Orthogonal Memory (AOM network) is used to provide a measure of the angle which form the 2 compared vectors too. In our system, we can appreciate that the method doesn't require an excessive processing time, so we can implement it for real time processing. Other advantage of the system is that it is applied to different types of clothes and defects (In general, other approaches are centred in only one type of defect). In the other hand, our algorithms produce rates of success around 94%. These results are quite encouraging if we keep in mind that it has been analysed some complex cloth types (such as lined cloth). To finish, and since the results obtained both in error rate and in execution times have been quite good, the application of this method can be very advantageous, moreover knowing that the development environment used is relatively simple.