Using Clustering and Metric Learning to Improve Science Return of Remote Sensed Imagery

Hayden, David S.; Chien, Steve; Thompson, David R.; Castaño, Rebecca

doi:10.1145/2168752.2168765

Cited by 6 publications

(2 citation statements)

References 38 publications

(31 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We have developed an technique, Multi-Domain/Multi-Class LDA (MDMC-LDA) and a corresponding regularization scheme which asllows LDA to exploit class structure local to an individual image while simultaneously capturing class relationships common to other images with related classes [12].…”

Section: Class (# Pixels)mentioning

confidence: 99%

Metric learning for hyperspectral image segmentation

Bue

Thompson

Gilmore

et al. 2011

2011 3rd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)

Self Cite

View full text Add to dashboard Cite

We present a metric learning approach to improve the performance of unsupervised hyperspectral image segmentation. Unsupervised spatial segmentation can assist both user visualization and automatic recognition of surface features. Spatially-continuous segments can improve noise levels and localization of feature boundaries. However, existing segmentation approaches rely on generic measures of spectral similarity. Here we learn domain-specific distance metrics based on training data, improving segment fidelity to semantic categories of interest. Multiclass Linear Discriminant Analysis produces a linear transform that optimally separates a labeled set of training classes. This defines a distance metric that generalizes to new scenes, enabling graph-based segmentations that emphasize key spectral features. We describe tests on data from the Compact Reconnaissance Imaging Spectrometer (CRISM) in which learned metrics improve segment homogeneity with respect to mineralogical categories.Index Terms-Segmentation, Metric Learning, CRISM HYPERSPECTRAL IMAGE SEGMENTATIONUnsupervised hyperspectral image segmentation can reveal spatial trends that show the physical structure of the scene to an analyst. They highlight borders and reveal areas of homogeneity and change. Segmentations are independently helpful for object recognition, and assist with automated production of symbolic maps. Additionally, a good segmentation can dramatically reduce the number of effective spectra in an image, enabling analyses that would otherwise be computationally prohibitive. Specifically, using an oversegmentation of the image instead of individual pixels can reduce noise and potentially improve the results of statistical post-analysis [1]. Typical segmentation methods for hyperspectral imagery include the watershed transform [2], Markov Random Fields [3], and the Felzenszwalb graph segmentation algorithm [4]. Generally speaking, they cluster pixels based on spatial proximity and a measure of spectral similarity given as a distance metric or other pairwise function. Existing hyperspectral segmentation approaches generally use generic distance measures that treat all channels equally or weight channels based on global statistical properties of the dataset. Such metrics are often confused by noise, instrument artifacts, or spectral variations that are irrelevant to semantic categories of interest.This work aims to improve automatic segmentations by learning a task-relevant measure of spectral distance from expert-labeled training data. We employ a multiclass Linear Discriminant Analysis (LDA) based approach to learn this measure. It produces segmentations that are not only are more visually cohesive, but also quantiatively more accurate in separating known materials into disjoint segments, in comparison to segmentations produced using the Euclidean metric. We evaluate this technique by comparing a set of expert-labeled mineral class maps to the segmentation maps produced by learned metrics, and provide a results on a case study focusing ...

show abstract

Section: Class (# Pixels)mentioning

confidence: 99%

Metric learning for hyperspectral image segmentation

Bue

Thompson

Gilmore

et al. 2011

2011 3rd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS)

Self Cite

View full text Add to dashboard Cite

show abstract

“…One of the emerging dimension reduction methods is distance metric learning (DML). Hayden et al [12] proposed a global DML algorithm, which maximizes the sum of all distances between samples from different classes and introduces two constraints to obtain an effective distance metric. Information theoretic metric learning (ITML) is a classic DML algorithm that transforms the optimization procedure of DML into a Bregman optimization problem [13].…”

Section: Introductionmentioning

confidence: 99%

Design and Application of an Image Classification Algorithm Based on Semantic Discrimination

Li¹,

Ge²

2019

View full text Add to dashboard Cite

Semantic gap is a common problem for most distance metric learning (DML) algorithms. Because of this problem, the semantic information may be inconsistent with the image features, which negatively affects the image classification accuracy. To solve the problem, this paper puts forward a new supervised DML method called semantic discriminative metric learning (SDML). The SDML maximizes the geometric mean of the normalized dispersion, making dispersions between different classes as identical as possible. Moreover, the Log function was combined with the geometric mean to further balance the dispersion between classes, and the maximum-margin criterion (MMC) was introduced to further enhance the discrimination ability of the distance metric. Finally, two constraints were applied to optimize the distance metric matrix. The effectiveness of the SDML algorithm was fully proved through experiments on actual datasets. The experimental results show that our algorithm outperformed many other typical DML methods in classification accuracy. This research provides an effectively way to measure the similarity between image samples and classify high-dimensional images.

show abstract