Deep Neural Network Methods for UAV Image Analysis
Farid Melgani (University of Trento, Italy)

Unmanned Aerial Vehicles (UAVs) are recognized as very effective systems for collecting images from a very low altitude. The high flexibility of these small, ecologic and silent aerial platforms permits immediate intervention and interactive measurements according to customer’s specific needs. Additionally, they allow mapping and monitoring of small areas at extremely fine scales and enables multitemporal acquisitions over the same area at predefined and desired times. These attractive proprieties render them a valid alternative or a complementary solution to satellite sensors particularly for small coverage or inaccessible areas. In the beginning, UAVs were used exclusively for military applications, but due to advancement in technologies and reduction in prices, they became a practical solution for many civilian applications. In this talk, we will describe different methodological solutions based on deep neural networks for urban monitoring issues.

---