Contents I. Introduction
Unmanned Aerial Vehicles (UAVs) are being used in many civil and military applications. The applied technology has been developed rapidly based on the integration of many technologies, control, communications, artificial vision [1]–[3], energy systems and mechanical systems. Such an interest is mainly driven by the low cost and its simplicity of operation. It is known that unmanned aerial vehicles are aerial position control systems [4]–[6] where the controller can have an off-line reference or an on-line reference. In the first technique, an algorithm analyzes a static scenery and then a vector trajectory (function trajectory) is generated based on the detected objects. Considering the vector as a reference, it is preloaded to the UAV flight controller to perform a displacement from the geographic point to another point is the latitude, is the longitude and z is the altitude. Some reported techniques are Waypoints [7] and Global Position System (GPS) combined with Robot Operating System (ROS) [8]. When these works are implemented it is expected that there will be no changes in the position of the objects or in the environment. For the on-line trajectory generation, a camera is placed on the UAV's surface, and this acts as an object sensing system [9]. With the information obtained from the video analysis, mathematical functions are generated and following, the optimal function is used as a reference in the UAV flight controller. The scene observed by the UAV is updated with each sensor sample and changes in the environment can be detected. To solve the trajectory generation problem, some authors propose different methodologies, for example: in reference [10], the authors use algorithms for segmentation, restoration, and mathematical methods to obtain information through characteristics of the video frames. In [11], the authors propose a route planning algorithm designed to work cyclically with the simultaneous localization and mapping (SLAM) estimation of a monocular-inertial system. It is based on keyframes and inertial measurement units (IMU). For the work reported in [12], the authors propose a direct visual servo motor system for a UAV, using an on-board high-speed monocular camera on a card with integrated GPU. It is characterized by recognizing the environment using artificial vision. However, the equipment is expensive (high-performance camera and embedded card) compared to other works, and it also requires prior knowledge or estimation of the object's spatial environment. The above cases are an example of the interest in achieving algorithms capable of generating optimal trajectories and that do not depend on prior knowledge about the scene.
