Hong Kong: Re­searchers de­velop AI-based fab­ric de­tec­tion sys­tem for tex­tile in­dus­try

Stitch World - - TECH TRACK -

The Hong Kong Poly­tech­nic Univer­sity has lately de­vel­oped an in­tel­li­gent fab­ric de­tec­tion sys­tem called ‘ WiseEye’, that uses smart tech­nolo­gies like Artificial In­tel­li­gence ( AI) and Deep Learn­ing in the process of qual­ity con­trol ( QC) in tex­tile in­dus­try.

The sys­tem ef­fec­tively min­imises the chance of pro­duc­ing in­fe­rior qual­ity fab­ric by 90 per cent, thus sub­stan­tially re­duc­ing the losses and wastages in pro­duc­tion. More­over, it also helps to save man­power as well as en­hances au­to­ma­tion man­age­ment in tex­tile man­u­fac­tur­ing.

‘ WiseEye’ is de­vel­oped by the Tex­tile and Ap­parel Artificial In­tel­li­gence ( TAAI) Re­search Team, which is headed by Prof. Calvin Wong from In­sti­tute of Tex­tiles and Cloth­ing, PolyU.

Prof. Calvin Wong ex­plained, “WiseEye is a unique AIbased in­spec­tion sys­tem that sat­is­fies the re­quire­ments of tex­tile man­u­fac­tur­ers. It is an in­te­grated sys­tem with a num­ber of com­po­nents that per­form dif­fer­ent func­tions in the in­spec­tion process.”

The sys­tem is sup­ported by AI-based ma­chine vi­sion tech­nol­ogy and can be in­stalled in a weav­ing ma­chine in or­der to help the fab­ric man­u­fac­tur­ers to de­tect the de­fects in­stantly in the pro­duc­tion process. With the help of au­to­matic in­spec­tion sys­tem, the pro­duc­tion line man­ager can eas­ily de­tect the de­fects, iden­tify the cause of prob­lems and fix them in­stantly.

It is em­bed­ded with high power LED light bar and high- res­o­lu­tion charge­cou­pled de­vice cam­era which is fur­ther driven by an elec­tronic mo­tor, mounted on rail to cap­ture the im­ages of the whole width of wo­ven fab­ric dur­ing the weav­ing process.

Fur­ther, these cap­tured im­ages are pre- pro­cessed and fed into the AI- based ma­chine vi­sion al­go­rithm so as to de­tect the fab­ric de­fects. All the real-time in­for­ma­tion gath­ered through­out the de­tec­tion process is for­warded to the com­puter sys­tem and an­a­lyt­i­cal statis­tics and alert can be gen­er­ated and dis­played as and when needed.

The re­search team has ap­plied ‘ Big Data’ along with ‘Deep Lean­ing’ tech­niques in ‘ WiseEye’. They have in­put the data of thou­sands of yards of fab­rics into the sys­tem and have trained it to de­tect about 40 com­mon de­fects with ex­cep­tion­ally high ac­cu­racy res­o­lu­tion of up to 0.1 mm per pixel.

Prof. Wong said, “In view of the nu­mer­ous fab­ric struc­tures that give great vari­a­tions in fab­ric tex­ture and de­fect types, au­to­matic fab­ric de­fect de­tec­tion has been a chal­leng­ing and un­ac­com­plished mis­sion in the past two decades.” He added that the in­no­va­tive in­tro­duc­tion of AI, Big Data and Deep Learn­ing tech­nolo­gies into ‘ WiseEye’ not only is a tech­no­log­i­cal break­through that meets the in­dus­try needs, but also marks a sig­nif­i­cant mile­stone in the qual­ity con­trol au­to­ma­tion for the tra­di­tional tex­tile in­dus­try.

Ac­cord­ing to PolyU, the sys­tem is still un­der trial for six months in a real-time man­u­fac­tur­ing en­vi­ron­ment. Also, the re­sults show that the sys­tem is able to re­duce 90 per cent of the loss and wastage in fab­ric man­u­fac­tur­ing process when com­pared to tra­di­tional hu­man vis­ual in­spec­tion. That means, it helps cut down the pro­duc­tion cost while en­hanc­ing pro­duc­tion ef­fi­ciency at the same time.

Prof. Wong and the TAAI re­search team have been con­duct­ing fun­da­men­tal and ap­plied re­search on

AI, com­puter vi­sion and ma­chine learn­ing, specif­i­cally for the fash­ion and tex­tile in­dus­try since 2012. The team ear­lier de­vel­oped ‘ Fash­ionAI Dataset’, which in­te­grates fash­ion and ma­chine learn­ing for sys­tem­atic anal­y­sis of fash­ion im­ages through the use of AI.

Ac­cord­ing to PolyU, the sys­tem is still un­der trial for six months in a real-time man­u­fac­tur­ing en­vi­ron­ment. Also, the re­sults show that the sys­tem is able to re­duce 90 per cent of loss and wastage in fab­ric man­u­fac­tur­ing process when com­pared to tra­di­tional hu­man vis­ual in­spec­tion.

Newspapers in English

Newspapers from India

© PressReader. All rights reserved.