Automatic Inspecting Machine System Of Fabric Defect Recognition Research And Application Of The Algorithm

In the textile process,there are some working procedures such as spinning.weaving.dyeing and so on.Cloth inspecting machine is an important part in the weaving process.Fabric defect detection is a key aspect of fabric quality control for the major textile mills. Currently,in the domestic textile industry,the defect detection of produced original cloth still remains in the manual inspection stage.Artificial cloth inspection machine has many weakness:slow.limited production,high undectected rate,poor detection continuity and so on.Consequently,there is an urgent need to develop novel,rapid and accurate methods for the automatic detection of fabric defects.The main difficulty of the fabric defect detection is that the texture and the morphology of the fabric is varied,especially the types of defects are vastly different, and therefore it is difficult to identify defects of all types.What’s more,detection speed subjects to certain restrictions because of the large amount of fabric image data processing.According to the problems about the varied types of fabric defects as well as a large amout of fabric image datas.it trys to use three level detection algorithm.As the transverse and longitudinal defects of the fabric occupy more proportion,the first level processing algorithm uses Hough transform to get the longest line for statistics for the number of pixels of linear,and then set a reasonable threshold according to the number of pixels for the first level detection.The second stage recognition algrithm uses the abnormal characteristics of gray projection waveform in the defect place to extract five effective features to recognise defects.The third stage recognition algorithm is a rapid location and recognition algorithm based on the gray projection and K nearest neighbor.First.it locates possible defect postion quickly based on the gray-level projecion. Second, it chooses a small rectangular area near the locating position as a research object.Then.it gets the features of the research object.Finally,it recognises the fabric defects by K neighbor algorithm.By offline simulation and online debugging results, it shows that this algorithm initially gets a good recognition effect. The simulation recognition rate is94%and the actual test result reaches76.8%.The result shows that this algorithm is stable.reliable,initially to meet the requirements of recognition performance and the speed of the algorithm.