Using thermal infrared detectors mounted on drones, and applying techniques from astrophysics, we hope to support the field of conservation ecology by creating an automated pipeline for the detection and identification of certain endangered species and poachers from thermal infrared data. We test part of our system by attempting to detect simulated poachers in the field. Whilst we find that we can detect humans hiding in the field in some types of terrain, we also find several environmental factors that prevent accurate detection, such as ambient heat from the ground, absorption of infrared emission by the atmosphere, obscuring vegetation and spurious sources from the terrain. We discuss the effect of these issues, and potential solutions which will be required for our future vision for a fully automated drone-based global conservation monitoring system.
Astro-Ecology couples ‘off the shelf’ infrared imaging technology and astronomy instrumentation techniques for application in the field of conservation biology. Microbolometers are uncooled, infrared systems that image in the thermal-infrared range (8-15μm). These cameras are potentially ideal to use for the detection and monitoring of vulnerable species and are readily available as ’off the shelf’ systems. However to optimise the quality of the data for this purpose requires thorough detector calibration to account for the systematics that limit readout accuracy. In this paper we apply three analogous, standard astronomical instrumentation techniques to characterise the random and spatial noise present in a FLIR Tau 2 Core thermal-infrared camera. We use flat fielding, stacking and binning to determine that microbolometer FPAs are dominated by large structure noise and demonstrate how this can be corrected by subtracting median stacks of flat field exposures.