An Improved SIFT Algorithm for Unmanned Aerial Vehicle Imagery

35th International Symposium on Remote Sensing of Environment

地球科学;生态环境科学

Li, J.M.^1,2 ; Yan, D.M.^1 ; Wang, G.^1 ; Zhang, L.^1

Key Laboratory of Digital Earth Sciences, Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing 100094, China^1

China University of Geosciences (Beijing), Beijing 100083, China^2

关键词: Computational speed;    Feature point extraction;    Gaussian image pyramid;    High resolution imagery;    Operational efficiencies;    Resources management;    Scale invariant feature transforms;    Topographic mapping;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/17/1/012187/pdf
DOI  :  10.1088/1755-1315/17/1/012187

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

The Unmanned Aerial Vehicle (UAV) platform has the benefits of low cost and convenience compared with satellites. Recently, UAVs have shown a wide range of applications such as land use change, mineral resources management and local topographic mapping. Because of the instability of the UAV air gesture, an image matching method is necessary to match different images of an object or scene. Scale Invariant Feature Transform (SIFT) features are invariant to image scaling, rotation and translation. However, the main drawback of a SIFT algorithm is its significant memory consumption and low computational speed, particularly in the case of high-resolution imagery. In this study, in order to overcome these drawbacks, we have analysed the construction of the scale-space in the SIFT algorithm and selected new parameters to construct the SIFT scale-space to improve the memory consumption and computational speed for the processing of UAV imagery. Here, we propose a restriction on the number of octaves and levels for Gaussian image pyramids. Our experiment shows that the proposed algorithm effectively reduces memory consumption and significantly improves the operational efficiency of the feature point extraction and matching under the premise of maintaining the precision of the extracted feature points.