An Autonomous Localization Approach For UAV In GPS Denied Environment

With the rapid development of UAV technology,UAV has been applied more and more widely,including military,agricultural,civil and commercial industries.The acquisition of accurate position and attitude plays a key role in UAVs’ navigation and localization.Currently,UAV navigation and positioning technology is a combination of inertial navigation and satellite navigation,such as INS/GPS.However,satellite navigation technology is usually vulnerable to interference and malicious damage,and there are some failure areas of satellite navigation.Too much reliance on satellite navigation systems will lead to a reduction in reliability and security of navigation.Without the support of GPS and other satellite navigation systems,a single inertial navigation system will not be able to achieve long navigation location because of the accumulated errors in the positioning of inertial devices themselves.The absence of position and attitude during UAV flight will cause great damage to its flight safety.Therefore,in view of the problem of UAV positioning under GPS rejection,this paper presents a vision-based global positioning technology.Firstly,the principle of vision localization is analyzed,and the relationship between the pose and real time image of UAV is studied through imaging system.The relative relationship and coordinate transformation of coordinates in visual location technology are studied through coordinate system,and further clarifies the principle of visual positioning by combining the technology of visual pose estimation.Secondly,from the point of view of registration,aiming at the registration problem of two images with small overlap area,the factors affecting the registration performance are analyzed.Based on the SURF detector and BRISK descriptor,an improved registration algorithm is proposed.At the same time,taking into account the fast speed of UAV and wide area of the flight area,the registration problem is further converted to the super large image registration problem.Aiming at the shortcomings of the existing super-large image registration,such as poor precision,long time and poor performance of the platform,this paper proposes a fast and high-precision super-large image registration algorithm based on overlapping regions on the basis of improved feature points algorithm.The experimental results show that the new algorithm can solve the limitations of the current processing platform,improve the speed and accuracy of the algorithm,and has better performance advantages than the existing algorithms.Finally,based on the multi-sensor information and prior knowledge,a global direct localization technology based on vision is proposed based on the super large image fast matching algorithm.This method estimates the approximate area and realizes on the fast matching algorithm.In order to improve the real-time performance of the algorithm,an improved global search technology is implemented.The improved positioning method improves the direct global positioning technology by adaptive fast search matching algorithm,which has better real-time visual positioning.Then,data fusion and correction of multi-source sensors,such as visual location information and INS data,are processed by Kalman filter to realize the high-efficiency positioning technology of UAV in GPS Denied Environment.