TESTING OF WOVEN FABRIC CONSTRUCTION USING PHOTOMETRIC STEREO TECHNIQUE WITH MULTI-IMAGE DISPLAY DEVICE

• Main Author: Setiawan, Irwan
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86555
• Institution: Institut Teknologi Bandung

Woven fabrics are widely used for various purposes. Woven fabric is obtained by weaving or crossing warp and weft yarns. Factors determining the quality of woven fabric include its construction. Testing the construction of woven fabrics still uses traditional methods and commercial testing tools, but it has many weaknesses. Computer vision methods have been widely developed to overcome problems in weaving construction testing. Still, current computer vision method research uses a lot of 2D (two-dimensional) fields that can be influenced by many factors and have low accuracy. The interweaving of warp and weft threads periodically produces a 3D (three-dimensional) fabric structure, so 3D surface structure analysis is needed. This research aims to develop a method for testing woven fabric construction, including linear mass, density, and weave patterns. The imaging method developed uses photometric stereo techniques with 3, 6, and 12 lighting variations with multi-image display devices and for stereo images using 3D surface structure analysis. In this research, hardware was designed and developed to record multi-view images of woven fabrics with various density variations, weave patterns, and images of threads taken from woven fabrics. Then, software development includes hardware control, 3D reconstruction of woven fabric images using unbiased photometric stereo techniques, and stereo image analysis for testing woven fabric construction parameters. Parameter testing of linear mass uses the empirical correlation equation between yarn diameter and linear mass. At the same time, yarn tethers are tested using integral projection, surface profile, and a Hough transformation and LOESS (Locally Estimated Scatterplot Smoothing) filter. Then, the basic woven fabric pattern parameters were tested using photometric stereo images, feature extraction was performed using wavelet image scattering, and classification was performed using PCA (Principal Component Analysis) and SVM (Support Vector Machine). Photometric stereo techniques with multi-image display devices can effectively test woven fabric construction. This research successfully developed a multi- image display device system with the optimum configuration: a lighting system with light sources = 12 tilt angle = 30o, light sources = six tilt angle = 60o, and light sources = three tilt angle = 120o. The slant angle = 54.74o. Then, the hardware for integrated image recording of yarn and fabric for weaving fabric construction testing was successfully developed and fabricated. In addition, software for hardware controllers, multi-view image reconstruction with unbiased photometric stereo algorithm, and stereo image analysis were successfully developed and made. In this research, multi-view image reconstruction using the unbiased photometric stereo algorithm successfully reconstructed 3D images of woven fabrics with 3, 6, and 12 lighting variations. Furthermore, linear mass testing using the empirical correlation method between yarn diameter and mass was successfully applied and tested on 100% cotton yarn, 100% polyester yarn, polyester and rayon blended yarn, and polyester and cotton blended yarn with an average error of 2.12%. In testing, the yarn density obtained an average error of 0.97%; this research has the advantage of being able to generalize the type of density so that it can be applied to all types of woven fabrics (dyed yarn, dyed woven fabric, and HTWF (high-tightness woven fabric)). While testing the weaving pattern using the wavelet image scattering method with PCA and SVM classification, it has an accuracy of 98.57%. The results of this study are expected to overcome the problems in testing woven fabric construction and replace the manual method of testing woven fabric construction in textile testing laboratories so that the use of photometric stereo techniques with multi-image display devices for testing woven fabric construction can speed up the testing process and improve measurement accuracy.