Hyperspectral remote sensing can improve the identification and classification of surface features through the spectrum comparing and matching to achieve classification and recognition. Because of the spatial resolution of the sensor as well as the difference in complexity and diversity on the ground, mixed pixels in the image are prevalent in remote sensing. The problem of subpixel unmixing is a prominent issue in the quantitative application of remote sensing. How to effectively interpret the mixedpixel is one of the key issues in the application of remote sensing. In the present paper, the hyperspectral reflectance characteristics of the mixedpixels formed with two kinds of materials whose area ratios have always been 1:1 were studied at different incident zenith angles and different topology location distribution, which provides a theoretical basis for the mixed pixel classification accuracy improve-ment.