Comparative study of panel and panelless-based reflectance conversion techniques for agricultural remote sensing

Small unmanned aircraft systems (sUAS) have allowed for thousands of aerial images to be captured at a moments notice. The simplicity and relative low cost of flying a sUAS has provided remote sensing practitioners, both commercial and academic, with a viable alternative to traditional remote sensing platforms (airplanes and satellites). This paper is an expanded follow-up study to an initial work. Three radiance-to-reflectance conversion methods were tested to determine the optimal technique to use for converting raw digital count imagery to reflectance maps. The first two methods utilized in-scene reflectance conversion panels along with the empirical line method (ELM), while the final method used an upward looking sensor that recorded the band-effective spectral downwelling irradiance. The methods employed were 2-Point ELM, 1-point ELM, and At-altitude Radiance Ratio (AARR). The average signed reflectance factor errors produced by these methods on real sUAS imagery were: -0.005, -0.0028, and -0.0244 respectively. These errors were produced across four altitudes (150, 225, 300 and 375ft), six targets (grass, asphalt, concrete, blue felt, green felt and red felt), five spectral bands (blue, green, red, red edge and near infrared), and three weather conditions (cloudy, partly cloudy and sunny). Finally, data was simulated using the MODTRAN code to generate downwelling irradiance and sensor reaching radiance to compute the theoretical results of the AARR technique. A multitude of variables were varied for these simulations (atmosphere, time, day, target, sensor height, and visibility), which resulted in an overall theoretically achievable signed reflectance factor error of 0.0023.