Skip to Content

Computer Vision for UAV, from visual information to visual guidance

Funded by the Spanish Ministry of Science MICYT #DPI2010-20751-C02-01

The main objective of this project is to provide next generation of UAV with the capability of robustly and efficiently using Computer Vision for the UAV control and guidance systems, as well as providing a rich processed visual information from the air, for a wide variety of civil applications. These civil applications of UAV in the non-segregate space will dramatically increase in the following years, accordingly to different sources, such as recent reports commissioned by the EC, the R&D programs of the most relevant Aeronautics companies and the international Aeronautics associations. Vision will play an important role in this matter of UAV integration in the non-segregate space for the whole range of UAV (from micro-UAV to commercial HALE). The technical matters that have to be solved and that are addressed in the current project can be clustered in the following generic topics, Robust Image Tracking, Visual Pose Estimation, Visual Servoing, Control and Guidance, Visual Sensor Fusión, Sense& Avoid Problem, and Aerial Image Recognition. These developed techniques are going to be applied in several micro an mini UAVs, of fix-wings, rotatory wings and quadrirotors, adapted for several applications. We aim to test the developed prototypes in several demos testbeds and also to participate in the two most relevant international competitions for UAVs, which constitutes a real challenge at international level to test the developed algorithms and platforms. 
The main objective is to be an outstanding international reference in the field of applying Computer Vision for UAVs. For this propose we plan to be the first Spanish team participating in the top international UAV competitions.

The technical objectives to be achieved are:

  1.  Development of new image acquisition systems: new acquisition devices will be explored, these systems will provide sufficient information (wide field off view, depth information) to improve the inspection and navigation tasks.
  2. Development of software architecture and control schemes for indoors and outdoors testbeds: the systems will be configured to tests the vision algorithms, this includes the development of control strategies to integrate the visual information with the UAV information.
  3. Improve robustness for visual tracking. This will be done using both vibration control and image processing algorithms.
    • In UAV’s some subsystems are sources of detrimental vibrations that significantly influence the mission performance, effectiveness and accuracy of operation of the vision system. A vibration-control systems as an isolator or an absorber will be included in the UAV. Moreover, a semi-active vibration control system will be implemented in order to reduce the amount of external power necessary to achieve the desired performance characteristics.
    • A software-based improvement of the image quality: image stabilization techniques.
    • Pyramidal tracking strategies:  pyramidal approaches in resolution and parameters (of the motion model), where at each pyramid level both the resolution and the motion parameters change.
  4. Sensor Fusion for pose estimation: the visual information will be used to recover the position and orientation of the UAV or an object of interest. This estimation will allow the development of visual-guided tasks. This objective include:
    • Onboard systems: pose estimation techniques will be developed, based on the information of onboard acquisition systems such as catadioptric, stereo, normal cameras.
    • External systems: pose estimation techniques will be developed based on information from multiple cameras.
    • New Sensor fusion strategies: to develop the vision-based control tasks and to improve the algorithms performance, robust methods to fuse the information from the different sensors (UAV's sensors and vision), are required. One objective is to propose new techniques to achieve the sensor fusion task.
  5. Visual navigation and flight control: as was exposed in the introduction of this section the main objective is to improve UAV capabilities by proposing vision-based solutions. These solutions include:
    • Autonomous take-off and landing: Vision strategies will be developed to solve these basic tasks. The nature of the tasks allow that the visual information can be useful to find safe places to land, to estimate the altitude and position of the UAV, to avoid obstacles during the tasks, among other characteristics
Starting date: 
Finishing date: