Research Of The Identification Arithmetic Of Fabric Defaults Using Computer Software

As a country with abundant textile production, we haven't exploited a practical automatic detecting system for textile yet. And there always been some problems in the systems imported abroad like: prices are too high, the configuration and operation of the device can't fit our actual manufacturing situation.Based on the study and research of the development of the automatic detecting technology of fabric defaults and the fabric defaults detecting production that already in use, we indicate the development direction of the research of automatic fabric defaults detecting of our country. That is the way to use computer software to detect. And we put forward a series of effective detecting arithmetic, which provides academic foundation and technological support for the own fabric defaults detecting system exploited by our country.The detecting method we used is listed bellow: take the photos of the white woven lighted by equal light with a numeral camera, and put the photograph of fabric into computer;first transform the fabric photo into gray photo, proportion the gray photo to enhance the definition and contrast;second use the method similar with the "gray level accretion matrix" arithmetic to identify the density of the fabric texture, compress the image data based on the texture density and the character of defaults, this expedite the subsequent processing speed and reduce the interference of noise;than use the gray value valve to divide the defaults from other normal fabric part and filter the noises by "corrupt-expand" arithmetic;last distill the eigenvalues of the default, input them to the single floor (no concealed floor) perceptron artificial neural network to classify the defaults.We apply the research production of fabric density identification into fabric default detection successfully, into the data compression of the fabric image process exactly. That will diminish the data account a lot, expedite the identification speed, and wipe off much redundant data, reduce the interference of noise, make the result more correct. In the experiment we detect 6 kinds of most constant defaults on thewhite woven: end out, double loop stitch, thread out, double loop pick, bore, oil stain, the correct identification rate reach to 96%.The method mentioned in this paper can be complement and optimize in the processes of: the texture density identification, data compression, the eigenvalue definition and distilling. The consummate arithmetic will got advance in processing speed, automatization level and variety of the defaults that could be detected.