The research on the UAV positioning method using the POS and improved image matching

Aiming at the problem of low accuracy of the target for one UAV(Unmanned Aerial Vehicle), according to the UAV’s characteristics of UAV aerial images, POS dates and satellite digital map, a UAV image matching and targeting method is presented based on POS. At first, using the Gaussian pyramid optical flow method, the UAV aerials or video frames are real-time stitched. Then, taking advantage of POS date, the approximate area of the entire aerials or video frames on the satellite digital map are determined based on the formulated search strategy. Finally, making use of the improved SIFT algorithm, the aerial photos(video frames) and satellite digital map in this range are matched to complete the targeting. The results proved that this method can achieve a wide range of real-time two-dimensional positioning without ground control points, while the precision and speed has large improvement, which is an effective method for target location utilizing image matching. 0 Introduction Currently, how to achieve positioning quickly and exactly for the goal is one of the core issues in the field of UAV research. The positioning accuracy of the traditional UAV map-matching and the atmospheric height measurements parameters method is under the influence of the underlay, low precision, and the application range is limited. In recent years, with the rapid development of the related computer vision algorithm performance and UAV’s POS technology, which makes it possible to use the image matching to achieve precise positioning, and the matching algorithm is the core content. Because Gaussian pyramid optical flow algorithm exists problems, many improved algorithms from different angles are proposed by scholars at home and abroad, mainly focusing in the generation process of simplification feature vector, which solve the matching problem of noise images and affine images. However, in the algorithm, the threshold value of the distance ratio can’t be adaptive, the matching accuracy is not guaranteed. In this paper, the feature that the UAV sensors can provide aerial imagery and POS information at the same time is used, as well as the advantages of a satellite digital map widely applied. Mainly for the two-dimensional positioning target, based on image matching, the UAV targeting method is proposed. without the condition of leading to the ground control points, using POS auxiliary aerial triangulation measurement and Gaussian pyramid optical flow matching method, the searching strategy is developed to complete adaptive positioning of different targets in a wide range, which plays important role for the fast and precise positioning of the UAV. 1 Position circuit summarization Launching the aerial photos or videos in the UAV shooting target area which is to be located, at the same time, obtaining POS data of the UAV at the moment. Received time: 2014-11-04 Fund Project: National Natural Science Foundation(51307183) About the Author: Zhang Yan(1991), man, Master, Mainly engaged in UAV information transmission and processing technology, E-mail:hillwind@126.com Corresponding Author: Li Jianzeng(1966), Man, Master Instructor, Mainly engaged in UAV information transmission and processing technology, E-mail:ljz681@sohu.com International Industrial Informatics and Computer Engineering Conference (IIICEC 2015) © 2015. The authors Published by Atlantis Press 1127 Firstly, using optical flow algorithm Gaussian pyramid method to stitch real-time UAV aerial photos or video frames and extract the overlapping parts. Secondly, making the use of POS data to aided aerial triangulation measurement to determine the ground objects three-dimensional coordinates of the overlapping part, and then calculate the ground objects two-dimensional coordinates of the entire aerial photos or video frames. According to the searching tactics developed in this article, finding the matching area on the satellite digital map. Finally, taking advantage of the Gaussian pyramid optical flow algorithm, the aerials or video frames are matched with the digital satellite map in this range, while getting the location information from digital satellite map to complete the targeting. The method uses the advantage of POS and satellite digital map, which is that getting location information quickly, achieving targeting real-time and exactly in a wide range, while avoiding the introduction of ground control points, the error influence of the orthographic correction and POS systems, which improves the targeting accuracy and speed, which is an effective targeting method using image matching based POS. 2 Gaussian pyramid matching method based on optical flow The degree of overlap of UAV video frame is large, the amount of motion between frames is small, so making use of optical flow method can register quickly and accurately. The Optical flow method approximately calculates motion field which can’t be obtained directly in image sequence through optical flow field, then estimating the relative motion between images according to motion field. This method is based on parametric modeling, which can describe the actual motion process better and avoid the problem of inaccurate motion vectors that is what hierarchical block matching method exists. Assuming that the motion field of the image f(x, t) with respect to the image f(xt, t)is p =(px, py)=x-xt. And px, py respectively represents the motion amount of horizontal and vertical directions. Motion field estimation makes the following sum of squared error Err reach minimum value: ( ) ( ) ( )2 τ Err f x, t f x p, t = − − ∑ (1) If the motion model is affine motion model: ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) 2 2 2 2 2 2 2 2 f x,t f x,t f x,t f x,t f x,t f x,t f x,t f x,t f x,t x y x y x x x x y x y x y f x,t f x,t f x,t f x,t f x,t f x,t f x,t f x,t f x,t x x xy x x xy x x x x y x y x y f x, y ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂                                     ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )