Constructing And Optimization Of Fabric Adptive Orthogonal Wavelet Based On Genetic Programming

Fabric defects detection and quality control are very important in woven fabric production process, but the traditional fabric defect detection mainly relies on artificial detection, which has a high rate of false detection and missed detection. Along with the wide application of computer technology, fabric defects detection, especially the fabric defect detection based on wavelet analysis is becoming the hot topics in recent time.As the diversity of fibers, yarn structures and fabric structures, fabric texture is very complex. It is very difficult to find a kind of wavelet device that can adapt to all fabric texture in order to sort out the numerous fabric textures. Therefore, we should firstly establish a wavelet library to list and stored up the orthogonal wavelet coefficients, and then to choose the wavelet adaptively based on the fabric features, so as to meet the needs of different fabric texture optimization.After constructing the adaptive orthonormal wavelets libraries, the article use the genetic programming algorithm to optimize the wavelet device matching fabric texture. Firstly, take the wavelet library as the population size of genetic programming algorithm, and optimize the control parameters of genetic programming algorithm. Secondly, we take the four common form of description of fabric texture including wavelet coefficients maximum minimum value difference, energy, fabric texture fluctuations and entropy as the fitness function of genetic programming algorithm. And the best fitness function is found after comparing the four fitness functions. Then take the normal fabric image as training samples to optimize the wavelet device matching fabric texture from the population size according to the genetic programming steps. Finally, decompose the defects fabric images by the wavelet. After extracting the characteristic value, binary and defect location operation, the fabric defects can be detected out, so as to realize the automatic test fabric defects. The conclusions getting from the subject are as follows. (1)The wavelet matching fabric texture can be optimized from the wavelet library by genetic programming algorithm combined with fitness function optimization method. After decomposing the defect fabric images by the wavelet, the defect information can be highlighted in the weft or warp direction sub images, and effect is better than the Monte Carlo method. The experiments proved that this method can effectively detect the common defects such as bamboo, double horizontal and weft shrinkage, thick weft, broken weft, lack of weft, double warp, hanging, warp-lacking, jump the yarn, grease, woven defect, hole defects and so on.(2)The fabric texture fluctuations are proved to be the best fitness function in the fitness function optimization of genetic programming algorithm. The wavelet getting by fabric texture fluctuations can be better match the fabric texture, stick out the defect information and weaken the background texture, so the fabric defect information can be showed in the binary image and the characteristic value.(3)It is better to extract the directional characteristic value when extracting the characteristic value of fabric images, that is, to extract weft value from the weft sub image, to extract warp value from the warp sub image. So it is easier to find the abnormal values of defect information in the fabric texture.(4)The method of separation of window location is a good way to find the positioning of the defect. It can locate the accurate range of fabric defect, especially when there is more than one kind of fabric defect or the defect distribution is in a wider range, this method can also locate the accurate range of the defect.