|Title:||Advanced textile image analysis based on multispectral colour reproduction|
|Advisors:||Xin, John (ITC)|
Hong Kong Polytechnic University -- Dissertations
|Department:||Institute of Textiles and Clothing|
|Pages:||xi, 98 pages : color illustrations|
|Abstract:||In this study, multispectral colour reproduction methods are investigated using textile images. Accurate colour reproduction of textile is crucial for the industry in terms of colour communication, colour measurement and quality control. Traditional colour measurement and colour communication are usually performed by spectrophotometers. The disadvantages of using spectrophotometers are that (1) they only measure a small area of the fabric at a time and (2) they lack the spatial reflectance information of the fabric.|
The contribution of this thesis can be listed as follows:
(1) Proposed an improved reflectance reconstruction method based on L1-norm penalization. Spectral reflectance reconstruction for multispectral images (such as Weiner estimation) may perform sub-optimally when the object being measured has a texture that is not in the training set. The accuracy of reconstruction is significantly lowered without training samples. Using L1-norm, our method can provide the transformation matrix with the favorable sparse property, which can help to achieve better results when measuring the unseen samples. We verify the proposed method by reconstructing spectral reflection for 4 types of materials (cotton, paper, polyester, and nylon) captured by a multi-spectral imaging system. Each of the materials has its own texture and there are 204 samples in each of the materials / textures in the experiments. Experimental results show that when the texture is not included in the training dataset, L1-norm can achieve better results compared with existing methods using colourimetric measure (i.e., colour difference) and shows consistent accuracy across 4 kinds of materials.
(2) Achieved accurate whiteness measurement in textile with the presence of fluorescent whitening agents (FWAs). Accurate whiteness measurement is important in many industries such as textile, paper and detergent production. With the advent of FWAs over the past decades, the general idea of whiteness has been changed from measuring reflectance to determining the spectral radiance factor (SRF) of the materials. Multispectral Imaging (MSI) system has long been investigated and is known to be an advanced technique to measure the reflectance of objects, by which one can use it to accurately measure the colour of objects. However, the spectral surface will be heavily augmented by the fluorescent materials added to the objects. As a result, measuring whiteness does not only depend on reflectance measurement but also SRF measurement. A specialized light-source is designed with an ultraviolet (UV) filter to control and adjust the illumination system in an integrating sphere. Through the adjustment of the position of the UV filter, multispectral images of an object (fabric in this work) are captured with different exposures of UV light from the light-source. Then these images are processed and combined by our proposed method into a new multispectral image with full-range visible spectral information. Besides, based on a theoretical proof, this work shows that MSI is spatial uniform for SRF measurement. Through the custom light-source with an adjustable UV filter, whiteness metric that is comparable with a spectrophotometer can be obtained by MSI. This work shows that MSI can achieve high whiteness measurement accuracy and spatial uniformity.
(3) Implemented a new clustering algorithm for measuring colours and recognizing patterns in printed fabrics. There are rich colours and various patterns with different sizes and shapes in printed fabrics, which make it difficult for accurate colour measurement and pattern recognition by traditional spectrophotometer and digital camera. This work proposes a grid-based density peaks clustering (GDPC) algorithm to measure colours and recognize patterns of multispectral image of printed fabrics. A custom-developed multispectral imaging system is used to capture the multispectral fabric image where each pixel has full spectral information across the visible spectrum from 400 nm to 700 nm with an interval of 10 nm. The multispectral image is then converted to CIELAB colour space for image processing (clustering) and colour measurement. The noise pixels are removed by calculating the local stability of each pixel, and then the remaining pixels are clustered by the proposed grid-based density peaks clustering algorithm based on the CIELAB colour values. Experiment results show that, when compared with conventional colour clustering algorithms, the proposed GDPC algorithm can have higher accuracy and efficiency in colour separation from multispectral images with complex patterns.
(4) Proposed a new approach to classify knitted fabrics. Automatic inspection of fabrics has tremendous advantage over the manual inspection due to its' great efficiency. Based on BoW (Bag of Word) feature extraction, a new approach for classification of knitted stitch was proposed. By this method, the classification of knitted fabrics can be significantly improved. To validate the method, we fabricated 58 texture knitted fabrics with 5 colours. The result shows our method can reach the best classification accuracy. This work will benefit the research of automatic recognition of textile pattern.
Overall, this work has improved the technical aspects in fabric industry using advanced computational methods.
|Rights:||All rights reserved|
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item: