Study Of Key Technologies For Scene Matching Aided Navigation System

In order to carry out the UAV navigation system, this paper mainly studied two important points: scene matching algorithm about multi-source image and the selection of the scene matching area are studied respectively.The good navigation matching algorithm which is precise, fast and can be applied to distance navigation is always the topic of the image matching field. Based on the character of edge, a multi-source image matching algorithm which performs good accuracy, adopt to UAV navigation system is put forward. This algorithm employed wavelet transform, MAD measurement and least square algorithm and had all the advantages which wavelet transform, MAD measurement and least square algorithm possessed. The algorithm of improved MAD similarity measurement is simple to compute and easy to realize in hardware. The pyramid structure of the image not only decreased the searching complexity, but also improved the speed of computing. In the upper layer, the improved MAD algorithm is used to find the candidate points which can ensure the veracity of the initial value in least square algorithm. In the lower layer, the better precise is achieved with least square matching algorithm and its computing complexity is compensate by The pyramid structure. It was showed that this algorithm is real-time and effective and meet the requirement of the UAV navigation system.A method of selecting the optimal scene matching area is provided and the storage of the reference-image is discussed. Analyzing the selecting rules, this paper proposed a hierarchical searching algorithm which is based on the characteristic of the original reference-image (the information of the image, the stability and the repeat mode). In this algorithm, the rule of selecting the matching algorithm in different layers is depending on the adoption of the matching algorithm.