Study On Scene Matching Navigation Technology Based On Real Time Image And Reference Image

Position and motion parameters are very important navigation date for the unpiloted aircraft. Traditionally these date acquire by the meaning of inertia and satellite navigation equipment. However, the error of inertia navigation equipment accumulates quickly as time going, and satellite navigation equipment has many restricts on our use, moreover it has poor resistance to interference. Because of measure accuracy and independence, Scene matching has become an important part of the hybrid navigation system, and attached more and more attentions.Scene matching is a process of matching real time image with reference image. Because of differences of scene, geometry and gray between real time image and reference image, Scene matching has low accuracy and precision, furthermore, scene matching can only provide position date for navigation system, can not provide velocity and direction date. These two factors are main faults in nowadays scene matching. This thesis has done some research into these two factors of scene matching, and has proposed the scene preprocessing which can eliminate the scene differences, the least squares image matching based on choosing points which has good real time performance and resistance of geometry distortion, the precise measurement of velocity and direction which is through scene matching in the whole time and calculating from many date. A lot of real time images scene matching experiments done through unpiloted aircraft prove that the performance of the scene matching navigation technology proposed by this thesis has reached the practice demand.This thesis is divided into three sections:Section I. This thesis has described some kinds of scene matching theories, and gave the methods how to realize them. Further more, this thesis has also pointed out the advantage and disadvantage of every theory.Section II. This thesis has analyzed the reason which made scene matching poor, and has proposed the common factor abstraction method which can improve the similarity between real time image and reference image, geometry emendation method which can eliminate the geometry distortion.Section III. This thesis has proved the apices stability principle in mathematic, and proposed the least squares image matching based on choosing points, which can resist geometry distortion and has good real time performance.Section IV. This thesis has proposed the accuracy measurement of velocity and direction based on scene matching for the unpiloted aircraft, and has analyzed the influence caused by the posture change of camera platform on measurement precise; furthermore, this thesis has proposed the method which can eliminate the error source.Section V. For the purpose of proving the method proposed by this thesis, we have done many scene matching experiments with the real time image acquired from unpiloted aircraft. The results of the experiments have demonstrated that the methods proposed by this thesis have good stability, precise and real time performance, which has reached the practice demand.