Comparison among filters and spectrometers in hyperspectral monitoring

Monitoring techniques in ‘proximal sensing’ coupled to image analysis insights are assuming an increasingly important role in a variety of discipline including ecology, forestry, geology, geography, urban planning, archeology, and civil engineering. In this context, the technique based on the acquisition and analysis of hyperspectral images appears the most suited for landscape classification and evolution over time. At the Laboratory of Fluid Mechanics of the Department of Civil, Construction and Environmental Engineering, an effective methodology for classifying landscapes according to the reflectance characteristics of the materials on the surface has been developed. It is based on the use of two complementary innovative experimental devices and appropriate software for acquiring and analyzing hyperspectral images. The first device is based on the use of three interference filters covering the ranges 400-720 nm (VIS), 650-1100 nm (NIR) and 850-1800 nm (MIR) and comprises one frame grabber with three Camera Link inputs and a trigger signal generator to tune the filter frequency, one solid state disk array of 1 Tbyte, two CMOS Dalsa-4M60 cameras (coupled to the VIS and NIR filters) and one Xenics InGaAs camera (coupled to the MIR filter), one thermal camera, a power supply for all devices and a processing computer for controlling the entire system and acquiring images of the thermal camera via a USB port. The second device employs the same equipments describe above but for the use of two spectrometers instead of three filters, the first spectrometer centered in the visible range (400 nm to 1000 nm, VIS), the second one centered in the near infrared range (from 900 nm to 1800 nm, MIR). The VIS spectrometer is mounted in front of the CMOS camera while the MIR one is mounted in front of the InGaAs camera. The other CMOS camera, equipped with standard lens, is used for georeferencing lines acquired with the spectrometers. In fact, the latter system simultaneously acquires the spectra of a line. Multiple images must be acquired to reconstruct a two-dimensional scene from the combination of several lines. The systems developed have very unique and qualifying characteristics, mainly related to the low cost if compared to other systems available on the market, high spectral resolution, high spatial and temporal resolution, easy portability, both systems have been engineered so that they can be transported by ultralight airplanes. It is suitable for monitoring areas of extension ranging from small to large. In the latter case a large number of images have to be acquired and joined into one large image without using ground control points. This can be automatically done through a specific software that calculates the correspondences by maximizing the photometric consistency among images through correlation operations. The software accounts for rotations, translations and object size changes in consecutive images. The platforms have been employed for field surveys aimed at acquiring hyperspectral images of objects placed at a distance of some hundreds of meters with the equipment mounted on the fixed stand (usually located uphill) or on aircrafts allowing low altitude flights. The most significant results have been obtained in the fields of vegetation recognition, detection of diseases and abnormalities in the spectral signatures of plant species (organic citrus on the Lentini Plains), identification and characterization of landfills, highlighting the spectral inhomogeneity they are associated with and the variations induced on land and on adjacent vegetation.