Abstract: The impurities rate is one of the important parameters to measure the quality of wheat combined harvesting mechanized operation. Obtaining impurities rate quickly and timely can grasp the quality of work of the combine harvester, which is very important for the agricultural production. In order to solve the problem of low efficiency of artificial detection of impurities rate and lack of mature grain impurities rate identification system in wheat harvesting mechanization operation, this paper tried to establish an spectral inversion model of impurities rate of wheat by using spectral technology, so as to achieve the goal of rapid nondestructive detection of impurities rate of wheat. The wheat spectral reflectance provided an alternative method to classical physical and chemical analysis of the impurities rate of wheat in laboratory. Therefore, the impurities rate of wheat was quickly achieved by using hyperspectral technology. First of all, totally 80 wheat samples were collected from the combine harvester model 4LS-7 made by Jindafeng. The impurities rate of these wheat samples was analyzed in the process of physical in laboratory. After that, the raw hyperspectral reflectance of wheat samples was measured by the FieldSpec 4 model Wide-Res instrument Made by ASD equipped with a high intensity contact probe under the darkroom conditions. Then, after preprocessing and mathematical exchange of the original spectral data,2 spectral parameters were obtained, namely, the original spectral reflectivity (REF) and the spectral reflectivity after logarithmic reciprocal treatment (LR). The impurities rate inversion model of grain combine harvester was established by using the 2 spectral parameters. Next, in order to get characteristic wavelengths, the application of principal component analysis (PCA) was explored in the optimization and quantitative analysis of hyperspectral bands. At the same time, the regression models of impurities rate with different parameters were established by least squares support vector machine (LS-SVM). Finally, the inversion results of the model were validated and compared with each other. The results showed that there were significant differences in the impurities rate of wheat samples obtained by mechanized harvesting, with the maximum and minimum impurities rate of 2.99% and 1.52% respectively. The mean impurities rate of the test samples was 2.28%, the standard deviation was 0.458, and the coefficient of variation was 20.09%. The experimental results showed that the sensitive bands of REF were 500, 689, 1 007, 1 117, 1 205, 1 211, 1 308, 1 381, 1 670 and 1 800 nm. Simultaneously, the sensitive bands of LR were 498, 502, 691, 700, 1 205, 1 373, 1 665, 1 788, 1 798 and 1 854 nm. The result indicated that PCA method could not only achieve the efficient selection of hyperspectral bands, but also retained the original sample information. The REF was the optimal spectral index in LS-SVM regression model (the modeling determination coefficient was 0.958, and the verification determination coefficient was 0.902). The REF hyperspectral inversion model based on LS-SVM can realize rapid monitoring of the quality of work of grain combine harvester in the future.