Visual feedback control of quadrotor by object detection in movies

In the previous report, the authors developed a quadrotor mission planning function using a smartphone and an autopilot function using GPS signals. In this paper, a visual feedback control is considered so that the system can automatically recognize the existence of an object in images or movies acquired by a quadrotor with a surveillance camera and this enables the quadrotor to track the object within the operational environment without using GPS signals. The visual feedback control is realized by automatic object recognition based on techniques of computer vision, in which the center of gravity position in the image coordinate system can be obtained from color or shape information. A visual feedback control application for flight control of DJI’s quadrotor Matrice100, and image analysis software are developed using Xcode and MATLAB, respectively. The validity and effectiveness are evaluated through experiments.