UAV-Thermal Camera Remote Sensing for Monitoring Wild Rabbit (Oryctolagus cuniculus) Population
MSc in GIS Dissertation Lenka Fronkova.pdf (4.904Mb)
MetadataShow full item record
This paper addresses challenges of remote sensing to detect, count and track wild rabbits (Oryctolagus Cuniculus) in their natural habitat. A combination of UAV, thermal and RGB cameras was used to survey the wild rabbit population. Tarot 680 Pro UAV caused significant rabbit disturbance and Phantom 2 caused low to medium disturbance. Due to the rabbit disturbance, the thermal camera did not record any rabbit images. Consequently, airborne remote sensing had to be changed to the rabbit remote sensing from the ground, utilising a tripod and manually focused FLIR A655 thermal camera. The ground system was successful in monitoring the wild rabbit, causing no animal disturbance. OpenCV Python computer vision library automated the noise removal, rabbit detection and count from the thermal imagery. However, this analysis was not fully automated, as parameters for each sequence had to be treated individually. The Lucas-Kanade sparse optical flow tracking algorithm improved the rabbit counts by preventing from double-counting the same individual. This analysis can be enhanced by applying Template Matching Binary Mask (TMBM), which classifies and finds objects of the same size, colour in the rest of the frames. This study shows that remote sensing of wild rabbits in a plan view using RGB sensor is problematic due to the rabbit camouflage against bare soil. Using a thermal sensor, based on the relative temperature, is feasible in the wild rabbit monitoring as it allows for recognition between hot objects and background. However, noise created by warm bare soil needs to be removed from the thermal imagery. A mast and a thermal camera or a quadcopter UAV and a thermal camera are recommended for the rabbit monitoring. These systems need a further testing. Similarly, the rabbit behavioural responses to different UAV types necessitate future research.