Bidirectional Reflectance Distribution Function Modeling Considerations in Small Unmanned Multispectral Systems

With the rapid advancement of miniaturized sensors and unmanned airborne system (UAS) platforms, scientific measurements that previously were carried out by large government or academic institution programs can now be successfully accomplished by smaller collection activities. In the case of vegetation assessment (e.g., forestry and agriculture), spatial, temporal, radiometric, and spectral resolution can be rigorously controlled. Although each of these is important, accurate temporal assessment of chemistry and morphological plant attributes is heavily dependent on spectral resolution and correction. A particular phenomenology of interest is the extraction and modeling of bidirectional reflectance distribution function (BRDF) data. In this effort, an approach is discussed to reflectance calibrate and derive BRDF signatures from soybeans through a newly available multispectral sensor integrated to an industry common UAS platform. Methods were developed to extract the reflectance data across the azimuth and elevation observations and fit these field data to previously derived models from the literature. A modular processing pipeline was developed to allow for the implementation of additional algorithms and efficient numerical analysis. Results show that the coefficients derived to fit the modeled BRDF data are consistent across spatial resolution and within cover type. Additionally, the modeled-fit root-mean-square error was inversely proportional to the spatial resolution of the image data used for signature extraction. In conjunction with more traditional spectral and ratio-based analytical indices, this approach provides important dimensionality in both classification and land-cover assessment applications critical to more accurate temporal vegetation assessment.