In this paper, Long Short-Term Memory, a special kind of recurrent neural network, is tested to search the feasibility of using machine learning approaches on the Terrain Referenced Navigation. By using the machine learning approach, we tried not to switch but to merge each strength of existing TRN algorithms; known as sequential, and batch processing methods; and make up their exclusive weaknesses. From our experiments, RNN structure originally designed to process natural languages, handwritten characters, and image captioning, also showed the possibility to learn how to read a terrain elevation map data to distinguish the current location of an aircraft. In this regard, we are trying to modify our network design to enhance the position estimation accuracy, and implement the system into real-world with the larger scale worldwide map data.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Terrain contour matching with recurrent neural networks


    Contributors:


    Publication date :

    2018-03-01


    Size :

    1051022 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    Contour matching: a curvature-based approach

    Seung Park, J. / Hee Han, J. | British Library Online Contents | 1998


    Contour Matching Based on Belief Propagation

    Xiang, S. / Nie, F. / Zhang, C. | British Library Conference Proceedings | 2006


    Hierarchical contour matching in medical images

    Xin, Y. / Truyen, B. / Pratikakis, I. et al. | British Library Online Contents | 1996


    Contour Matching for 3D Ear Recognition

    Chen, Hui / Bhanu, Bir | IEEE | 2005