Vision Aided Localization And Navigation For Unmanned Aerial Vehicle In GPS-denied Environment

At present,unmanned aerial vehicles(UAVs)are the focus of research in the of maritime warfare.UAVs must complete a series of tasks such as take-off,navigation,and landing in a variety of complex mission environments.The landing phase is a high-failure stage.Therefore,it is imperative to study a safety and autonomous landing system for UAVs.Because the visual navigation system has the advantages of low power consumption,small size,non-contact,easy concealment,and strong anti-interference ability,this paper proposes a visual-aided UAV autonomous landing navigation system in the GPS-denied environment.This system is based on monocular vision technology to acquire the three-dimensional position and attitude angle of UAV,and based on Kalman filter,visual navigation,integrated navigation of INS and radar altimeter,INS cumulative error correction and ship speed information pre-estimation are realized.This system based on the monocular The main research contents of this paper are as follows:1)First of all,the landing target is designed in shape,size,color and material,so that it has the characteristics of both long and near distance,direction guidance and all-weather nature and so on.In addition,in order to improve the real-time performance of the system,the camera model is studied and carried out the off-line calibration.2)This paper proposes an image preprocessing scheme for remote distances in RGB color space and near distance in HSV color space,aiming at the different characteristics of landing target,which reduces the computational complexity of subsequent visual processing algorithms.3)In order to identify the safe landing platform identified by the landing target during the autonomous landing process of the UAV.In this paper,a recognition scheme based on the combination of freeman chain code and geometric features of landing target is proposed when the long distance cooperative target is fully visible,and the ellipse recognition scheme based on AMIs when the close cooperative target is not fully visible is proposed.4)As the ship is in motion,in order to prevent the loss of the target,a corner tracking method based on the pyramid LK optical flow method is proposed.In the close range,the color features of the cooperative target are used,and the CAMShift tracking method based on the color histogram is adopted,and the automatic switching of the two tracking algorithm is realized,which improves the real-time performance of the visual navigation system.5)In order to solve the problem of pose solving in visual navigation systems,several coordinate systems involved in the system and their transformation relations are studied.In order to achieve the one-to-one correspondence between 3D feature points and their 2D projection points,a feature extraction algorithm based on DP algorithm is proposed,and the a feature point sorting algorithm based on the geometric features of cooperative objects is proposed an algorithm for sorting the precise target points of the landing target is proposed.An Ellipse Four-point extraction algorithm based on Elliptic fitting and Elliptic projection model is proposed at close range.Finally,the PnP algorithm is combined with the camera’s internal parameters to understand the pose and position the UAV.In order to improve the accuracy and robustness of navigation system,two kinds of sensors,INS and radar altimeter,were introduced.A multi-rate information fusion scheme based on Kalman filter is proposed,which makes full use of sensor information in the system,and the velocity state was introduced into the state equation of Kalman to predict the speed of the ship.The error accumulation model of INS is studied,and an INS error correction scheme is proposed based on the model and Kalman filter.In order to verify the navigation system proposed in this paper,a 3D demonstration scenario was designed using real-time 3D virtual reality development technology.The state of the ocean was statistically analyzed,and its impact on the ship was analyzed,and the movement parameters of the existing ships were studied.A more realistic ship motion model is created.The INS and radar altimeter were simulated with reference to the accuracy of the navigation level INS and the accuracy of the radar altimeter.Finally,the results of the demonstration verification system are analyzed.In real-time,the visual navigation system can achieve a frame rate of at least 17 frames to meet real-time requirements.In terms of accuracy,the position accuracy is 0.5m,the attitude angle accuracy is 1.5?,and the velocity accuracy is 0.8m/s.The feasibility of the visual navigation system and the data fusion and INS correction scheme presented in this paper is proved.