The recent developments in the performance and miniaturization of unmanned aircraft systems (UAS) and multispectral imaging sensors provide new tools for the assessment of the spatial and temporal variability of soil properties at sub-meter resolution and at relatively low costs, in comparison to traditional chemical analysis. The accuracy of multispectral data is nevertheless influenced by the anisotropic behaviour of natural surfaces, framed in the general theory of the bidirectional reflectance distribution function (BRDF). Accounting for BRDF effects in multispectral data is paramount before formulating any scientific interpretation. This study presents a semi-empirical spectral normalization methodology for UAS-based multispectral imaging datasets of bare soils to account for the effects of the BRDF, based on the application of an anisotropy factor (ANIF). A dataset of images from 15 flights over bare soil fields in the Belgian loam belt was used to calibrate a model relating the ANIF to a wide range of illumination geometry conditions by using only two angles: relative sensor-pixel-sun zenith and relative sensor-pixel-sun azimuth. The employment of ANIF-corrected images for multispectral orthomosaic generation with photogrammetric software provided spectral maps free of anisotropic-related artefacts in most cases, as assessed by several ad hoc indexes, and was also tested on an independent validation set. Most notably, the standard deviation in the measured reflectance of the same georeferenced point by different pictures decreased from 0.032 to 0.023 (p < 0.05) in the calibration dataset and from 0.037 to 0.030 in the validation dataset. The validation dataset, however, showed the presence of some systematic errors, the causes of which require further investigation.