On the Co-Registration of Asynchronous Multi-Spectral and Thermal Images

Abstract

Abstract. Plastic pollution has a severe impact on the ecosystem, altering its natural equilibrium and causing serious health issues to both flora and fauna. Several actions have already been undertaken in order to reduce the plastic litter dispersion in the environment, both in terms of changing the human behavior, reducing the use of plastics and avoiding their dispersion, and of implementing methods for detecting and collecting the already dispersed ones. This paper focuses on the latter, and, in particular, on plastic litter detection on the fluvial environment. To this aim, an Unmanned Aerial Vehicle, provided with a multi-spectral and a thermal camera, have been used, in order to: (i) allow affordable periodic monitoring of relatively long river reaches, (ii) detect even quite small macro-plastics, based on their spectral signature. More specifically, since the cameras deployed in our data collection campaigns are not synchronized, this work aims at presenting the developed strategy for the co-registration of the acquired imagery, which results to be quite challenging given the few amount of visual features recognizable on the images acquired flying at a limited altitude over a river. The proposed methodology, which is based on the correlation maximization between multi-spectral and thermal images, provided reasonable results on the considered case study. The obtained values of normalized intersection over union of plastic areas are over 80%.

Monitoring the spread of a pathogenic insect on vineyards using UAS

“…The second step is calculating the camera exterior orientation using Agisoft Metashape. Starting from the NIR images collected, a sparse point cloud is generated, and through the collimation of at least three of the stable points defined in T0, it is possible to obtain more accurately the NIR camera exterior orientation parameters, since the drone mounting the Sentera sensor was not equipped with accurate navigation sensors (GNSS receiver with Real Time Kinematic corrections and accurate Inertial measurement Unit) (Cortesi et al, 2023, Teppati Losè et al, 2023. The third and last step of the CV-based method is the spatialisation of the detected insects based on the reprojection of their estimated locations in the vineyards.…”