- SEMINARIO - Applications of Computer Vision to Drones - Line Following Control, Human Detection and Safety Classification of Landing Sites

DISERTANTE: Dr. Ing. Felipe Nascimento Martins, del Instituto Federal de Educación, Ciencia y Tecnología de Espirito Santo, Brasil.
RESUMEN: Drones (or Unmanned Aerial Vehicles - UAVs) are versatile machines that can be used in a variety of applications, such as automatic monitoring of crops and water channels, pest detection, animal counting, search and rescue etc. Autonomous flying is a desirable feature for drones, especially for those that are frequently used in monitoring and inspection of large areas. In this presentation we are going to present two applications of computer vision to drones. The first one is the use of features of the environment as references for the navigation control of the drone. This is important because in some situations global positioning system signal is not guaranteed or its error might be too large, hence other methods of local position feedback are required. We present the development of a vision-based line following strategy to guide an autonomous UAV to follow water channel margins, crop lines and other similar patterns, to support automatic monitoring and inspection activities. We present the design of a nonlinear path following controller and we show that the resulting closed-loop system is stable in the sense of Lyapunov. The second application is a system to be used in drones to make search and rescue operations more effective. The system uses a drone downward facing camera to detect people and to evaluate potential landing sites as safe or unsafe. Histogram of Oriented Gradients (HOG) features are extracted and a Support Vector Machine (SVM) is used as classifier. Some experimental results are going to be presented and discussed.