Efficient and Effective Hierarchical Feature Propagation

Abstract

Many methods have been recently proposed to deal with the large amount of data provided by the new remote sensing technologies. Several of those methods rely on the use of segmented regions. However, a common issue in region-based applications is the definition of the appropriate representation scale of the data, a problem usually addressed by exploiting multiple scales of segmentation. The use of multiple scales, however, raises new challenges related to the definition of effective and efficient mechanisms for extracting features. In this paper, we address the problem of extracting features from a hierarchy by proposing two approaches that exploit the existing relationships among regions at different scales. The H-Propagation propagates any histogram-based low-level descriptors. The bag-of-visualword (BoW)-Propagation approach uses the BoWs model to propagate features along multiple scales. The proposed methods are very efficient, as features need to be extracted only at the base of the hierarchy and yield comparable results to low-level extraction approaches.

A Comparative Study of Sampling Analysis in the Scene Classification of Optical High-Spatial Resolution Remote Sensing Imagery

“…In recent years, numerous studies have been devoted to the scene classification of optical HSR-RS imagery; see, e.g., [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17]. Although the methods of scene classification involve many different methodologies, such as the bag-of-visual-words (BoVW) models [1,4,9], the topic models [3,8] and unsupervised feature learning-based methods [10,13,14], it is clear that one can summarize these methods into three main components, namely patch sampling, feature learning and classification, each of which has a considerable influence on the final performance.…”

“…Scene classification of optical HSR-RS imagery, which aims to classify extracted subregions of HSR-RS images covering multiple land cover types or ground objects into different semantic categories, has recently attracted considerable attention in remote sensing image interpretation [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17].…”

“…Observe that, in order to do scene classification for a large-scale HSR-RS image, one often requires to first construct the basic scene elements for classification, by using either non-overlapping tiles [2,5,13], overlapping tiles [10,14,16] or super-pixels [4,11]. However, the core problem is how to establish the scene models, i.e., exploiting the variations in the local spatial arrangements and structural patterns inside local tiled regions.…”

Foreword to the Special Issue on Pattern Recognition in Remote Sensing

“…In [6], the bag-of-visual words (BOVWs) model is applied to high spatial-resolution (HSR) image classification and categorization; specifically, a concentric circle-structured multiscale BOVW method using multiple features is proposed, which is superior to many existing BOVW methods in solving land-use scene classification problem. In [7], feature extraction from a hierarchy of segmented regions is studied for multispectral image classification; the bag-of-visual-word-Propagation approach propagates features along multiple scales, which are very efficient and can yield comparable results to low-level extraction approaches. In [8], morphological profiles (MPs) are considered for classification of HSR hyperspectral images, where multiple structuring elements (SEs) with different shapes are proposed to use because they together can produce higher classification accuracy with spatial-spectral information.…”

Learning to semantically segment high-resolution remote sensing images

“…Accordingly, many works have discussed the advantages of region-based classification against the classical pixel-wise approach. However, classification considering segmentation-based techniques is still an open task [5], [6]. The main challenges are related to the semantic aspects inherent to the extraction of objects of interest, which is directly related to the selection of the most suitable segmentation scale.…”