An evolution of fabric inspection technology in textiles
Garment rejection/alteration due to weaving or knitting defects in the fabric can lead to a huge value loss to the company. Fabric inspection is considered to be one of the most difficult of all textile processes to automate. In this article, authors (Dr Rashmi Thakur, Dr Deepak Panghal and Dr Prabir Jana) track the evolution of fabric inspection technology in textiles from manual to AI-based system. Read on...
Fabric has always remained the center stage in garment industry, it accounts for the maximum cost, takes maximum lead time to procure and change its texture and colour every season to attract the consumer while causing headache for production personnel. Fabric inspection has proven to be one of the most difficult of all textile processes to automate. Garment rejection/alteration due to weaving or knitting defects in the fabric can lead to a huge value loss to the company. Not only the defect identification process, but also the location marking of the defect on the fabric and avoidance of the defect while marker making and subsequently in cutting, is of immense significance so as to offer a defect-free garment.
Manual fabric inspection
At preliminary level fabrics are inspected visually on a horizontal or slanting inspection table illuminated from the top. For greige fabric and for some special fabric types additional illumination from bottom also provided, so that the defects are better visible to naked eye. While the slanting table offers better ergonomics than horizontal table, both can be custom manufactured in house at manufacturer end. Flat table is preferred for fabric packages that come in meter fold, while slanting tables are better for roll type of package.
A fabric inspector either flips the fabric (for meter fold type package) or pulls the fabric (for roll type package) over the inspection table. Once a defect is found, the inspector pastes a small paper sticker next to the defect or make a circle with a contrast colour chalk (to be easily identifiable in subsequent process) and record the defect specification on a paper inspection form or through a digital screen.
Mechanised fabric inspection
Mechanisation of the manual process involves motorised unwinding and winding of fabric, digital or analog length measuring device, fabric edge guiding device, and start-stop mechanism by inspector. The fabric inspection machines are power driven with variable frequency drive inverter for precise speed control and proximity sensors to control the uniformity of edges, with an auto-stop option. The machine speed can vary between 12 meters/minute to 60 meters/minute and some models have provision for the option to inspect delicate fabrics at zero tension to avoid damage to the structure of the fabric.
To document defect pattern for analysis and decision making, ‘fabric inspection defect analysis software’ (FIDAS) is available in the market which can be installed on every fabric inspection machine of any brand and make. The defect can be entered through a touch screen monitor and performs automatic fabric gradation based on number of defects. The software works on the 4-point system for fabric inspection as per ASTM standards and available client server based as well as web-based system. Some of the solution providers are Datalog, Inspectrum, etc.
Some of the other features available are edge alignment device, rollers for creaseless movement of fabric, stretch control mechanism, de-curling system for knits, roll hardness/compactness adjustment, roll ejection mechanism, autostop at feed, if the machine runs out of fabric and most importantly port to FIDAS. Some of the most popular Indian brands of fabric inspection machines are Ramsons, Paramount, Gayatri Engineers, Almac Group, etc.
Over the years the technology has improved for existing processes (like mechanical edge alignment replaced by laser guided, etc) as well as many new features were added in the fabric inspection machine. While the speed increased, consistency improved and physical exhaustion reduced, the importance of human inspector remained the same. The defect is still identified by a human inspector through naked eye and defect type is categorised by a human brain.
Automatic camera-based fabric inspection
Automated human-less fabric inspection systems are based on adaptive, neural networks which can learn. These fabric inspection machines are equipped with CCD/CMOS cameras, which scan the fabric using advanced Fractal scanning techniques. Fractal scanning technique was necessitated because of typical construction of woven fabric where warps and wefts are interlaced at 90 degree angle and density of yarns can exceed 100 per inch in any direction. Conventional raster scanning was found insufficient as there is a possibility of a defective yarn completely missed the scanning frame. Fractal scanning technique ensured all defects big or small, warp or weft is being captured.
Most of the first generation of automated fabric inspection technologies used spectral approach for identifying defects from camera image as this technique can only be used for materials having a high degree of periodicity in its texture and not for materials containing random texture. With spectral analysis methods, an image is converted into spectral domain using a suitable orthogonal transform, e.g, the Fourier transform, the Gabor transform, and the orthogonal wavelet transform. Although the technology suppliers did not reveal the exact methods used in their machines, it is assumed that they used Fuzzy Wavelet Analysis. Although the solution has been there for decades, the acceptability was limited to large textile producers with low variability in product range and also prohibitive cost.
The image acquisition device may consist of an array of fixed cameras and a LED illumination bar or one or two moving image acquisition heads, consisting of a camera and illumination unit. While scanning across the fabric, images are taken and transferred to the image processing unit. Here, proprietary algorithms are applied to analyze the texture of the fabric and to detect deviations from standard.
Some of the brands offering solutions in this space are Cyclops and Argus, WebSPECTOR by Shelton Vision Systems, IQ-TEX 4 by Elbit Vision Systems.