Research On Aerial Image Location Based On Image Matching

Scene matching navigation compares and matches the aerial image captured by the aircraft in real time with the map data pre-stored,then the current location is obtained according to the geographical location information carried by the map data.Because of its high accuracy and autonomy,scene matching navigation is of great value as an auxiliary navigation technology.The pre-stored image that participate in matching have different source from the aerial images in scene matching,and often are complex natural scenes with low resolution,and there are problems of matching time-consuming caused by massive pre-stored images.The core of scene matching is the choice of image matching algorithm,so the key problem to be solved is the real-time matching between heterogeneous images.This paper intends to use the scene matching to achieve the positioning of the aerial image according to the actual needs.Aiming at this problem,this paper proposes a solution based on the actual and different image matching algorithms.The whole matching process is divided into two parts: rough matching and exact matching.The main work of this paper can be summarized as the following three aspects:1.Proposing rough image matching algorithm.Rough matching algorithm is intending to quickly extract the image with the highest similarity with the aerial image in the pre-stored image,so as to reduce the number of subsequent exact matching.Firstly,two preprocessing methods are proposed to reduce the impact of heterogeneous images on matching results: based on saliency feature extraction and KFCM clustering,respectively.Then,for the local overlap problem,this paper selects the gray-based template matching method to measure the similarity between the real-time image and the pre-stored images.2.Improving the SURF algorithm in exact matching algorithm.The SURF algorithm used for exact matching has superior performance,but the whole matching process takes too long.Therefore,this paper proposes a SURF feature extraction method based on saliency feature weighting,which realizes the feature extraction feature only on significant area of the image to reduce the number of feature points.Secondly,for the mismatching situation,this paper combines the two methods of ratio detection and homography estimation to achieve the deletion of mismatch.3.Optimizing the integrated solution.In addition to the improvement of the algorithm,this paper prompts two methods to reduce the running time in the process of program execution combined with practical considerations.Firstly,the results related to the pre-stored image needed in the whole process can be calculated in advance and stored as pre-stored data.It can be directly called in the actual matching eliminating the processing time for pre-stored images.Secondly,taking into account the time-consuming and execution order of each step,parallel computing can reduce the running time.Finally,summarize the methods proposed in this paper,as well as the problems in the method.The solution is proposed for the existing problems,and the future research contents and research directions are prospected.