The MultiSpectral Image Analysis and Understanding research line is focused on exploring the use of spectral information complementary to the visible spectrum to develop advanced computer vision systems. Using multi-spectral acquisition systems, the goal is to respond to difficult-to-solve vision problems using only conventional cameras. The main research topics of the group are the design of acquisition systems, image registration, 3D reconstruction, non-rigid model matching, semantic segmentation, super-resolution and the detection and classification of objects. These problems are attacked by techniques of global optimization and deep computational learning.
Although the research is transversal, the area of application where the group mainly focuses is the analysis of multi-spectral aerial images, obtained either with satellites or drones. In this context, the group works in precision farming, the characterization of forest fires and the geolocation of elements of interest.