Study And Implementation Of Position And Attitude Estimation For Four-rotor Mini Rotorcraft Based On Vision

Since Four-rotor Mini Rotorcraft has simple configuration, low fabricative cost and many other merits, it has been put into comprehensive uses in military domain, astrospace exploration, disaster rescue, geographic information surveying, terrain mapping, environment surveillance, agricultural and forest applications, electrical power line examination, film production industry, and many other areas. Hence, a lot of universities and research organizations all over the world now pay their attention on the mini aircrafts as the study focus in Unmanned Aerial Vehicles (UAV). But the study of the basic theory and the experimental applications on the Four-rotor Mini Rotorcraft is still at the initial stage. In the study, the fusion of many technologies is involved, and among them the navigation control is one of the key technologies in UAV. Moreover, one of the advanced navigations for the Four-rotor Mini Rotorcraft is about visual system.The visual information system loaded in mini aircraft can be applied to observe environment, avoid obstacles, identify and track objects. It can obtain position and attitude data, angular velocity data, speed data, and other data of the flight from the images taken by the system. These data provides flight parameter information which is needed by its flight control system. With this system, some problems in the autonomous navigation flight of mini aircraft under complex environment of near ground may be solved.In this dissertation, the Four-rotor Mini Rotorcraft is taken as the study object. The study starts from establishment of the mathematical model of position and attitude of the rotorcraft, and the estimation method of the position and attitude of its flight based on vision is discussed. An image processing algorithm for the position and attitude estimation of this rotorcraft for engineering implementation is proposed, and the flight control system and the vision-based position and attitude estimation system are designed and completed.All work involved in the dissertation is as the flowing:The first chapter provides a summary of the research background and significance of the mini aircraft. The study and application on the visual position and attitude estimation of mini aircraft both at home and abroad is introduced. Then the study goal and contents of this dissertation are given.The second chapter focuses the Four-rotor Mini Rotorcraft as the study object and its visual position and attitude parameter estimation as the study content. Based on the analysis of the flight control mode which is novel and special for the Four-rotor Mini Rotorcraft, the mathematics model of the visual position and attitude control is established for rotorcraft. It provides the basic theory for the control of position and attitude of the rotorcraftt. According to vision technology with the geometry perspective projection, a theoretical method of the position and attitude parameter estimation of vision-based by camera installed in the aircraft is further studied and discussed when the aircraft is stably in hovering. Thismethod can provide the theoretical reference in the designing of the hardware and developing software for the implementation of visual position and attitude estimation system of the rotorcraft.The third chapter proposes the arithmetic frame for the visual image information processing based on the design of the artificial landmark for the visual position and attitude estimation. By the analysis of experimental results from the visual image processing in the laboratory, an adaptive threshold segmentation method of dynamic statistic combined with primary colors is proposed for dynamic binary pre-processing on the landmark images captured by the camera in aircraft, and the method provides the better target images information for later visual image processing. After the basic analysis of the characteristics of the artificial landmark design, the projection method is applied to divide multi-objective image characteristics region of the visual position and attitude estimation into small areas. And the method combined with both projection and wavelet transform to get the central elliptic coordinates information from target images in the small areas partitioned is presented, and thus it can provide the necessary information of the points coordinate for the visual position and attitude estimation.In the fourth chapter, by explanation of the basic structure and principle of the vision system in mini aircraft, and based on the design requirements of visual position and attitude estimation, two feasible design schemes are discussed. After the comparison of these design schemes, a more suitable scheme for the implementation is chosen, and the entire system framework on four-rotor mini rotorcraft system is given, including a visual system and a flight control system. According to the system framework, design and implementation of the hardware and software platform on the rotorcraft's vision system is provided. And the system is successfully applied to the stable flight tests of the rotorcraft in hovering. The experiment results have shown that the relative precision error of the parameter estimation of the visual attitude estimation system on the rotorcraft is within 8.5 degrees.In the final chapter, the brief summary of the work conducted in this dissertation is given and some further research problems are proposed.The main innovations in this dissertation are as the following:A mathematical model of the visual position and attitude parameter control on the Four-rotor Mini Rotorcraft is established, which provides a theoretical foundation for the control of flight position and attitude. Based on the relations of the geometric perspective projection, the theoretical approache of the position and attitude estimation by using visual image information from the aircraft camera is studied. It provides a theoretical reference for the design, development, and implementation of both software and hardware of the four-rotor min aircraft.Under the laboratory environment, the adaptive threshold segmentation method of dynamic statistic combined with the primary colors is proposed, which is used to pre-process the image by using dynamic binary processing. On this basis of the above method, a theorem that on a coordinate in the maximum peak of the elliptic projection is the coordinate of the elliptic centre is provided. Then this method combined with both projection and wavelet transform is proposed to quickly divide the image in multi-object areas and get the coordinates information of the elliptic centre in each small areas. The real-time processing capability of the visual position and attitude estimation in embedded DSP system is increased by using the image processing algorithm for engineering implementation.By using color CCD camera in the aircraft and multimedia video processor, the rotocraft can carry out image acquisition and quantification of the landmark for the visual position and attitude estimation system. Since the color information is applied to the visual position and attitude estimation system, the information quantity on identification of the character of object image is improved.A solutions to the Four-rotor Mini Rotorcraft system is proposed by the combination of the embedded DSP visual system and the flight control system. The design and implementation of the software and hardware platform for the vision of the Four-rotor Mini Rotorcraft is given. And the flight of the rotorcraft in hovering is tested under laboratory conditions.