The supervised learning method used in an on-line optical transmission mid-near inspection of the coating weight of basis weight fabrics for a textile finishing are investigated. The real-time scanning width supervised learning method and area-based strategy for determining based of process quality of textile finishing manufacturing. The modified supervised learning for least-squares method is used to obtain the suitable parameter transformation between the measured voltage and the nonwoven fabrics basis weight. The supervised learning least-squares method obtained as the parameter transfer equation. We consider textile finishing basis-weight inspection error amount is less than 2%. Can provide more accurate and effective basis weight verification information and to practitioners for their factory applications.