Under the optimal test conditions explored in the early stage

the online original near infrared spectrum for the polyester/cotton blend fabrics samples was collected using a self developed “high efficiency identification and sorting device for the main components of fiber products”.Based on the original online spectrogram

an optimal spectral pretreatment method was proposed as S-G smoothing+MMN+S-G derivative.Meanwhile

an online near infrared quantitative analysis model for polyester/cotton blend fabrics was established by partial least square method.The root mean standard error of cross validation(RMSECV) was 1.47

the correlation coefficient of calibration(RC) and correlation coefficient of validation(RV) were not less than 0.99

the relative predictive deviation of calibration(RPDC) was 18.17

the relative predictive deviation of validation(RPDV) was 13.13

and the relative predictive deviation of cross validation(RPDCV) was 1176.To verify the reliability of the model

30 external samples were selected for online verification.The linear equation for the verified results was y=(1.00±0.01)x-(0.88±0.56)

with an accurate rate of 93.3%.After importing the model into the “textile online master control program” of the sorting device

the sorting category of the equipment was set for different polyester content fabrics

then the content prediction for the polyester/cotton fabrics samples were performed

and the samples were automatically purged into the corresponding collection frame through the device's blowing and sorting system.The time for each sample to be predicted and sorted was less than 2 s

and the automatic mechanical sorting result was correct.Therefore

the established model and sorting device could be used for efficient online measurement and automatic sorting of waste polyester/cotton blended textiles.