Study On Scene Aided Location Method For Ground Target Based On UAV Image Sequences

Accurate and real time positioning has always been a problem that needs to be solved urgently in the development of military power.In the most common use of inertial navigation and other angle measuring device to obtain the position information of the aircraft is a direct positioning method.In the typical UAV military reconnaissance environment,in the face of the lack of ground control information,light and small UAV can not equipped with high precision inertial navigation system and due to the need the ground control point to correct the error of attitude drift this typical method is difficult to apply.Therefore,it is very important to study the method of real time and high accuracy of the ground control point based on the sequence image analysis.Through the image of the ground to achieve positioning needs to solve the problem of pose angle of the aircraft,using of multiple view geometry theory in computer vision from the same image points to meet the basic matrix we can obtain camera position and the relative position between the camera to achieve the reconstruction of scene in a similar transformation.On this basis,the 3D position of the real scene can be obtained by eliminating the similarity transformation.In this paper,for the actual situation of different scenes,we focus on the positioning method which based on the implementation of the scene reconstruction,and the case of known ground height to study the height constraints to improve the positioning accuracy of the method.The main work done includes:Firstly,the paper introduces the advantages and disadvantages of the existing methods,and analyzes the advantages and disadvantages of various positioning methods.For the general situation,different from the traditional method of measuring the aircraft attitude is to equip with the attitude measuring device on UAV,in this paper,we study the relationship between the two views and according to the basic matrix to obtain the relative position and attitude between the cameras,we studied the 7-points algorithm,linear method and robust solution,and the results are compared with the results of different estimation methods.According to the decomposition of the basic matrix,the camera matrix is recovered,and the reconstruction of the common scene is realized without any other auxiliary information.Taking into account that the large area of the ground surface can be used to be a planar scene in the real UAV reconnaissance environment.In view of this kind of plane scene,this paper studies the planar homograph of the two views,and introduces the calculation method and the decomposition theory,we studied reconstruction of planar scene based on the decomposition of the homograph matrix.In the UAV reconnaissance mission of no ground control information and the attitude measurement device,the aircraft position information is known but with the impact of a certain error,this paper studies the scene reconstruction aided to the the UAV positioning method,the location of the ground is further realized based on the reconstruction results of the ordinary scene and the plane scene,using the traditional method of bundle adjustment method,the optimization of the aircraft attitude and points 3D-coordinates is used to ensure the positioning accuracy.In addition,in order to get the high accuracy of the results,this paper studies the method of the location of the ground under the height constraint,under the influence of different noise and different simulation conditions,the comparison of the positioning results with the height restriction is achieved,the results show that with the height constraint can achieve high accuracy georeferencing.