Elements Extraction Of Apparel Patterns Based On Contrast And GrabCut Algorithm

The intelligent design of apparel products is an important part of the high-quality development for the apparel industry.Facing a bulk of clothing patterns,the rapid and efficient reading of the pattern information still has the following problems:First of all,the clothing pattern uses the fabric as a carrier,the inherent texture in the fabric will interfere the pattern information and affect the effectiveness of the pattern content;moreover,clothing patterns often contain multiple pattern elements.Pattern elements contain independent internal meanings,which is the key to pattern information extraction.The traditional method of manually extracting elements of patterns has obvious shortcomings when processing large batches of images.Thirdly,the existing automatic extraction process of elements of clothing patterns often only considers the outline information,and ignores the color information and internal structure of the pattern,which cannot achieve effective storageIn view of the above problems,the research is carried out on the following two aspects:removing the texture from the garment pattern image,retaining the effective information and automatically extracting and storing the elements of the pattern.The main contents of the paper are as follows:This paper analyzed the research status and development trend of the digitalization of clothing patterns.Then carried out the research with the printing fabric patterns as an example.100 clothing pattern images were obtained with the scanning software VIZOO x-tex.A software platform for the processing of clothing pattern images was built:Open CV 3.4+VS 2017.In order to improve the transmission efficiency,bilinear interpolation is chosen to normalize the size of garment pattern.The bilateral filtering algorithm is used to remove the texture of the carrier fabric,while the main structure,color and edge contour information of the pattern are retained.Compared with square filter,mean filter,gaussian filter and median filter,it is proved that our method is superior.In order to highlight the color pattern information,the adjustment of pattern contrast and brightness is added in the experiment.For the shortcomings of the GrabCut algorithm that relies on prior knowledge entered manually,the HC and AC saliency analysis methods are introduced to obtain salient and non-saliency regions to separate the pattern foreground and background.It is used as the prior knowledge of pattern segmentation to input GrabCut algorithm.Initially optimize the full covariance Gaussian mixture model(GMM)and realize the final segmentation of color clothing pattern elements.The method in this paper can reduce the amount of manual interaction and improve the calculation speed.From the perspective of user operation,the design and development of an integrated program interface for the extraction and storage of clothing pattern elements has been completed Then it was applied to the clothing pattern element extraction example,randomly selected 30 clothing pattern elements,and compared the results of manually extracting the pattern elements from the PS.The average recall rate of the proposed algorithm was 99.426%,and the average accuracy rate was 98.703%The average value of Jaccard similarity index is 98.139%,and the average value of F-measure comprehensive index is 99.016%.It seems that the extraction effect of basic elements in this paper can meet the design requirementsThis subject proposes a method suitable for the extraction of the main structure of a color pattern image,which can remove the texture structure of the fabric tissue while retaining a large amount of pattern edges,detailed structures,and color information.This paper proposes a pattern element extraction method based on contrast and GrabCut algorithm,which eliminates the parameter setting link and requires only a small amount of user interaction to extract clothing pattern elements with "shape,color,and structure",which can achieve fast and effective storage.The proposed method can make the pattern design and production links more intelligent and efficient,and also provide a certain reference for the development of clothing design auxiliary systems.This subject proposes a suitable method that extracting the main structure of a color pattern image,which can remove the texture structure of the fabric tissue,and can retain a large number of pattern edges,detailed structures,color information.The methond can improve the quality of the pattern.A pattern element extraction method based on contrast and GrabCut algorithm was proposed,whitch could eliminate parameter setting step,just need small amount of user interaction in extract clothing pattern elements.Our algorithm can keep the pattern shape,color,structure,and enable fast and effective storage.The proposed method can make pattern design and production more intelligent,and also provide a certain reference for the development of clothing design auxiliary systems.