Abstract: Near-ground imaging spectroscopy applied in field provides new opportunity for development of quantitative remote sensing in agriculture. It deserves concern about how to exert its data advantage of integrating image and spectra into one, particularly in analyzing the influence of background targets, such as soil, shadow on crop nutrient inversion model. In this research, imaging cubes of wheat group in the field were collected by visible/near-infrared imaging spectrometer (VNIS). A normalized spectral index was set up according to reflectance characteristics of illuminated soil, shadow soil, illuminated leaf and shadow leaf in the image. Furthermore, the index was used to extract spectra of different targets in soybean images and analyze the variation of determination coefficient R2 between normalized spectra of soybean group and chlorophyll density before and after removing background soil. The results showed that when spectra of soil and shadow leaf were removed, the sensitive bands of chlorophyll density shifted from red and near-infrared ranges (727 nm, 922 nm) to red ranges (710 nm, 711 nm), meanwhile, the overall trend was that visible ranges increased, near-infrared regions decreased and red bands had the highest determination coefficient. Therefore, it can be concluded that spectral purification based on normalized spectral index has important significance for quantitative research in agricultural remote sensing.