Unmanned Aerial Vehicles: Embedded Control

Chapter 12

Real-Time Stabilization of an Eight-Rotor UAV Using Stereo Vision and Optical Flow ¹

Image-based control to stabilize at hover the rotary wing machine is performed by separately applying two different techniques of computer vision. In Chapter 9 a single camera has been used to estimate the aerial robot position: now in this case two cameras are used to estimate the unmanned aerial vehicle (UAV) position, i.e. applying the stereo vision theory (epipolar geometry) the 3D position is obtained. Separately another computer vision technique is developed to estimate indirectly the UAV translational velocity on plane x − y. This technique is based on optical flow, which calculates the image velocity between two consecutive frames (characteristic pixels displacement). With these optical flow measures, the vision system is able to make a very acceptable estimation of flying machine translational speed. Similarly as in Chapter 9, the vision system is used only to estimate the position and velocity of UAV, while the attitude states (angles and angular speeds) are in charge of the IMU.

The small aerial vehicle used in this experiment is an original configuration having eight rotors. Four rotors are devoted to stabilize the orientation of the helicopter and the other four are used to drive the lateral displacements [ROM 09]. The proposed configuration is particularly useful for image processing since the camera orientation is held constant. The eight-rotor rotorcraft is simpler to ...