Reconstruction Of Textile Texture Structure And Hyperspectral Imaging Analysis

Textile appearance detection and textile component testing are important components of textile testing,its test indexes determine the quality of textiles,and guide the production process and production methods.However,traditional methods of artificial detection are time-consuming and inefficient,and it is difficult to accurately and accurately record the appearance and composition of textiles.Therefore,in view of the textile appearance test textile texture structure reconstruction method is proposed in this paper,using the fabric multi-angle image sequence synthesis of 3 d model of an outline of the fabric,measurement,characterization will provide a basis for the follow-up;To compose a textile component detection is proposed in this paper highlights like analysis technology,the hyperspectral images of fabric is acquired after the analysis of images,to establish a suitable model for fabric according to the components are classified,in order to realize the fabric composition test.This paper from the two aspects of textile appearance inspection and textiles ingredients outlined the main detection methods at home and abroad in recent decades and detection method,and the shortage of the traditional methods are put forward.Secondly introduced the three dimensional reconstruction and compose a specular highlights as the development of the technology background and the present application in all walks of life,according to the characteristics of the textiles put forward a new technology is introduced into the advantages and innovations in the field of textile testing.In this paper,according to the characteristics of the textiles to built a set of multi-angle image acquisition platform,by using single camera angles for fabric surface texture image 9,after the calibration of the camera to obtain the camera internal and external parameters and spatial location,and then used to extract the feature points matching image information more complete accurate matching,finally use point cloud image filling,texture mapping,complete fabric surface model reconstruction,to reconstruct the model is analyzed,and the results show that the model by the overall accuracy is higher,but in view of the woven fabric more hairiness or count higher reduction effect remains to be improved.In addition,a high spectral identification system for fiber components including hardware equipment and software support platform is also built.More than in the eight kinds of pure textiles group collection fabric after the hyperspectral image,after the hyperspectral image preprocessing,image according to the ratio of 1:3 according to the composition of divided into calibration and validation sets.Then,the feature bands of each image were extracted using the continuous projection algorithm(SPA)and principal component analysis(PCA),and the classifier was established based on the least squares(LS-SVM)and the nearest neighbor algorithm(KNN).Finally,the two methods are used to extract the calibration set feature wavelength into the classifier for training,and the classifier that completes the training is used to classify the verification set.Classification results show that the recognition rate by using two methods in more than 98%,which is based on the continuous projection algorithm and least squares support vector machine(SPA-LS-SVM)classifier is better,the recognition rate is 100%.The results show that the reconstruction of textile texture and the analysis of the high light of textiles have high accuracy,which can replace traditional textile appearance detection and component detection methods.