Detection of Non-Uniformity in Dyed Textile Materials Using Image Analysis

Fig A explains the proposed schematic flowchart of fabric defect identification process using Artificial Neural Network(ANN). Fabric images will be captured using CCD cameras and several contrast (CON) values, which will be used as input parameters to ANN model, will be extracted corresponding to various directions and pixel distances. Then the ANN model will be trained using the image features as the inputs. The output layers of ANN model will haves nodes to identify different types of faults. The incorporation of other image features like Angular Second Moment (ASM) can also be tried to improve the efficacy of the system.

FigureA : Flowchart of fabric defect classification using image analysis and ANN

In an effort to make the textile dyeing industry more robust, researchers at IIT Delhi have collaborated with the Vardhman Group for devising a technology that will improve the end product.

The project titled ‘Detection of Non-Uniformity in Dyed Textile Materials Using Image Analysis’ is a part of the Uchchatar Avishkar Yojana (UAY) of the Ministry of Human Resources (MHRD).

India has been one of the world's largest manufacturers of textiles and garments. Abundant availability of raw materials and a skilled manpower helped India clinch a position of a sourcing hub. It is the world’s second largest producer of textiles and garments.

Though initially, India had concentrated on lot of investments in spinning and weaving, slowly it became apparent that it is in the Dyeing/printing sector where the key value addition in the fabric manufacturing chain lies.

“One of the major problems with the dyeing machineries still lie in uneven dyeing. There are several reasons for the faults generated – non-uniformity in raw materials, improper preparation of fabric prior to dyeing, uneven pressure of rollers, unequal heating during drying. However, till date all such faults are majorly detected offline, often causing huge losses for the industry. So the motivation was to find a faster system, which is not subjective and which would result in a lesser number of rejections from the buyers,” says principal investigator Samrat Mukhopadhyay.

The project would, with the help of the industry partner aim to develop a fabric detection system which would detect and analyze unlevel dyeing. The image acquisition, analyses of defects, analyses and classification of the defects using suitable algorithms would be developed as a part of the project.

A suitable image acquisition system would be developed, which would capture images at reasonable clarity for analyses. An algorithm would be developed (a) to convert colored images into grey scale (optional) and (b) increase contrast. A suitable flowchart would be devised based on analyses from different parts of the sample to detect uneveneness. Different modes would be used to detect short term and long term variations.

The main deliverables from this project would be a system to detect fabrics with unlevel dyeing. It is also envisaged, that the reason behind the unlevel dyeing would be ascertained to a reasonable degree of accuracy.