The Design And Implementation Of Four Rotor UAV Fixed Landing System Based On Machine Vision

Four rotor UAV has the advantages of low cost, remote control, and can be used for vertical takeoff and landing, which can fly over the area, which can be widely used in the commercial field. To achieve UAV’s intelligent and autonomous flight, so that the UAV can automatically perform aerial cable patrol task,is an important application field in power grid enterprises. UAV autonomous designated habitat technology is the basic technology to achieve autonomous UAV flight. With the help of Autonomous Fixed Landing Systems, UAV can autonomous land on the high-altitude building, self-charging, achieve the goal of continuing to operate autonomously.This paper uses Open Source Computer Vision (OpenCV) library,to design and implement the Fixed Landing System.basing on machine vision, the Fixed Landing System is running in the "unmanned aerial vehicle (UAV)/server" mode. The system is based on four axis aircraft flight principle, by use of C/C++ application development, OpenCV open source vision library development, image saving and processing, software engineering and related technologies, so that the UAV can achieve the efficient operation of the landing tasks.In this paper, the main work and contributions are as follows:1) Designed the "unmanned aerial vehicle (UAV)/server" mode architecture of the system. This system using the wireless video transmission protocol, the drone captured video transmits to the server, with the ability of the server operation real-time calculation unmanned aerial vehicle (UAV) and identifies the relative position, direction control of unmanned aerial vehicle (UAV).2) Used the Open Source Computer Vision library for data processing. Processing scheme including color image gray level transformation, image filtering, image binarization, enhanced code readability and maintainability of the system, greatly shorten the development cycle of the system.3) Used the ellipse fitting method for visual target tracking location. For each has been image binarization, system efficient point of focus from the inside first find all can find out the closing of the outline object. For each object, the system for an ellipse, make it as much as possible consistent with the object. Namely, for object point set, with elliptic equations for the model fitting, make a point with the object point on the elliptic equations set overlap, the parameters of the elliptic equations are obtained, used to indicate the outline of the object. Finally determine the ellipse center, determine the target. Ellipse fitting can very quickly find circular graphics, and can quickly find center, suitable for various occasions of fixed-point habitats.This paper provides an accurate and real time implementation scheme for the UAV patrol line. Its performance meets the needs of practical application, and achieves the design goal.