hyperspectral images of 47 unit wheat seed samples are collected

and average spectrum information is extracted within the range of interest. Combined with three index basic data of moisture

protein and wet gluten content reflect the quality of wheat seed

partial least squares（ PLS） quantitative analysis models of the three index are built separately the optimal spectral preprocessing and waveband combination is searched by OPUS software.And by comparing the hyperspectral models with the NIR models

it is found that the performance index of hyperspectral model is better than that of NIR model. The results show that

when the sample is granular and its chemical composition inside distributes unevenly

the accuracy and stability of the NIR model will be affected by the limitation of its measurement conditions

while hyperspectral technology exhibits larger sampling area

more comprehensive information

and stronger analysis potential

which provides a new way for wheat quality evaluation.