用于大型遥感影像检索的深度学习.pdf
需积分: 10 49 浏览量
2020-04-10
22:18:56
上传
评论 1
收藏 1.83MB PDF 举报
1
Deep Learning for Image Search and Retrieval
in Large Remote Sensing Archives
Gencer Sumbul, Jian Kang, and Beg
¨
um Demir
Abstract
This chapter presents recent advances in content based image search and retrieval
(CBIR) systems in remote sensing (RS) for fast and accurate information discovery from
massive data archives. Initially, we analyze the limitations of the traditional CBIR systems
that rely on the hand-crafted RS image descriptors applied to exhaustive search and
retrieval problems. Then, we focus our attention on the advances in RS CBIR systems
for which the deep learning (DL) models are at the forefront. In particular, we present the
theoretical properties of the most recent DL based CBIR systems for the characterization
of the complex semantic content of RS images. After discussing their strengths and
limitations, we present the deep hashing based CBIR systems that have high time-
efficient search capability within huge data archives. Finally, the most promising research
directions in RS CBIR are discussed.
F
1 INTRODUCTION
With the unprecedented advances in the satellite technology, recent years have witnessed a significant
increase in the volume of remote sensing (RS) image archives. Thus, the development of efficient and
accurate content based image retrieval (CBIR) systems in massive archives of RS images is a growing
research interest in RS. CBIR aims to search for RS images of the similar information content within a large
archive with respect to a query image. To this end, CBIR systems are defined based on two main steps:
i) image description step (which characterizes the spatial and spectral information content of RS images);
and ii) image retrieval step (which evaluates the similarity among the considered descriptors and then
retrieve images similar to a query image in the order of similarity). A general block scheme of a CBIR
system is shown in Figure 1.
Traditional CBIR systems extract and exploit hand-crafted features to describe the content of RS images.
As an example, bag of-visual-words representations built upon the scale invariant feature transform fea-
tures are introduced in [1]. In [2], a bag-of-morphological-words representation of the local morphological
texture descriptors is proposed in the context of CBIR. In [3], a comparative study that analyzes and
compares advanced local binary patterns (LBPs) in RS CBIR problems is presented. To define the spectral
information content of high dimensional RS images, the bag of spectral values descriptors are presented
in [4]. Graph-based image representations, where the nodes describe the image region properties and the
edges represent the spatial relationships among the regions, are presented in [5], [6]. In [7], [8] image
representations through binary hash codes are introduced for large-scale CBIR problems in RS. In detail,
in [7] kernel-based hashing methods are presented, whereas in [8] a partial randomness hashing method
is introduced.
Once image descriptors are obtained, one can use the k-nearest neighbor (k-NN) algorithm, which
computes the similarity between the query image and all archive images to find the k images most similar
to the query. If the images are represented by graphs, graph matching techniques can be used. As an
example, in [6] an inexact graph matching approach, which is based on the sub-graph isomorphism and
spectral embedding algorithms, is presented. If the images are represented by binary hash codes, image
• Gencer Sumbul, Jian Kang, and Beg ¨um Demir are with Technische Universit¨at Berlin, Berlin, Germany. Corresponding Author:
Beg¨um Demir; demir@tu-berlin.de
arXiv:2004.01613v1 [cs.CV] 3 Apr 2020
资源评论
