Researchers have introduced a unique drone prototype to help search and rescue teams locate missing persons – even in dense forests. Subjects in individual thermal images often appear completely or partially hidden; the drone instead combines several individual images into one integral image (image integration) that can be used for classification and to better detect people.
Image integration by airborne optical sectioning is a novel technique for creating high-resolution images of large-scale scenes from a moving platform. It uses a laser scanner to illuminate a thin section of the scene and a camera to capture the reflected light. Combining multiple images from different viewpoints and orientations can reconstruct a 3D scene model with fine details and textures. This technique has remote sensing, mapping, surveillance, and cultural heritage preservation applications.
As image integration significantly reduces concealment, modern deep-learning methods can aid in correctly detecting individuals up to a probability of well over 90%. The same procedure attains a recognition rate of less than 25% when using conventional individual images. As there is little to no data available to support this kind of classification, the team had to create its training database over the past few months.
Initial field study findings and pivotal early insight indicate that image integration significantly supports classification purposes better than individual images when it comes to concealed objects.